The surge in large language models (LLMs) has fundamentally reshaped the landscape of GPU usage patterns, creating an urgent need for more efficient management strategies. While cloud providers employ spot instances to reduce costs for low-priority (LP) tasks, existing schedulers still grapple with high eviction rates and lengthy queuing times. To address these limitations, we present GFS, a novel preemptive scheduling framework that enhances service-level objective (SLO) compliance for high-priority (HP) tasks while minimizing preemptions to LP tasks. Firstly, GFS utilizes a lightweight forecasting model that predicts GPU demand among different tenants, enabling proactive resource management. Secondly, GFS employs a dynamic allocation mechanism to adjust the spot quota for LP tasks with guaranteed durations. Lastly, GFS incorporates a preemptive scheduling policy that prioritizes HP tasks while minimizing the impact on LP tasks. We demonstrate the effectiveness of GFS through both real-world implementation and simulations. The results show that GFS reduces eviction rates by 33.0%, and cuts queuing delays by 44.1% for LP tasks. Furthermore, GFS enhances the GPU allocation rate by up to 22.8% in real production clusters. In a production cluster of more than 10,000 GPUs, GFS yields roughly $459,715 in monthly benefits.

Knowledge base question answering (KBQA) aims to answer natural language questions using large-scale knowledge bases (KBs). Among various KBQA approaches, semantic parsing-based (SP-based) methods have demonstrated strong effectiveness by generating concise logical forms (LFs) that capture complex subgraph structures and semantic information. Recent research suggests that integrating large language models (LLMs) with SP can achieve significant improvements in the performance and efficiency of KBQA by facilitating the direct generation of LFs with minimal retrieval. However, generating complete LFs with LLMs continues to pose a challenge due to the complexity of the required graph structures and constraints, leading to the significant issue of non-executability. To address these challenges, we propose GCA-KBQA, a step-wise fine-tuned LLM-based framework that employs hop-wise generation, knowledge-assisted calibration, and path-level assembly to construct complete LFs for KBQA. Specifically, we decompose the complex SP process into manageable steps: first, we iteratively generate LFs for each topic entity one hop at a time using a fine-tuned LLM, leveraging KB knowledge to calibrate intermediate outputs and mitigate error propagation. Subsequently, we guide the LLM in assembling path-level LFs from different topic entities, resulting in optimized final LF. We evaluate the proposed method on four KBQA benchmarks spanning two distinct KBs, demonstrating its superior performance compared to state-of-the-art baselines. The code is available at https://github.com/pvfeldt/GCA-KBQA.

Detecting anomalies in business processes is imperative for achieving operational success, especially as multiple participants increasingly engage in collaborative efforts to complete processes, i.e., collaborative business processes, amidst rapid economic development. However, existing business process anomaly detection approaches are typically designed for centralized training, making them impractical for collaborative business processes that require stringent privacy measures. In this paper, we introduce a feDerated Unsupervised AnomaLy detection framework for collaborative business processes, named DUAL. DUAL enables participants to collaboratively detect anomalies via a third-party coordinator, which constructs a global view from the hidden representations of participants' local sub-traces without exposing the sub-traces themselves. To further strengthen privacy protection, DUAL injects differential privacy noise into the transmitted hidden representations. To mitigate the global-local discrepancy, we introduce novel event execution errors which indicate whether the client who is supposed to execute the reconstructed event is consistent with the observed one. Extensive experiments demonstrate that DUAL effectively detects anomalies in collaborative processes while preserving privacy, achieving performance comparable to a centralized setting that requires access to participants' raw sub-traces.

Detecting anomalies in business processes is crucial for ensuring operational success. While many existing methods rely on statistical frequency to detect anomalies, it's important to note that infrequent behavior doesn't necessarily imply undesirability. To address this challenge, detecting anomalies from a semantic viewpoint proves to be a more effective approach. However, current semantic anomaly detection methods treat a trace (i.e., process instance) as multiple event pairs, disrupting long-distance dependencies. In this paper, we introduce DABL, a novel approach for detecting semantic anomalies in business processes using large language models (LLMs). We collect 143,137 real-world process models from various domains. By generating normal traces through the playout of these process models and simulating both ordering and exclusion anomalies, we fine-tune Llama 2 using the resulting log. Through extensive experiments, we demonstrate that DABL surpasses existing state-of-the-art semantic anomaly detection methods in terms of both generalization ability and learning of given processes. Users can directly apply DABL to detect semantic anomalies in their own datasets without the need for additional training. Furthermore, DABL offers the ability to interpret anomalies' causes in natural language, providing valuable insights into the detected anomalies.

The end-to-end automated design of machine learning (ML) pipelines significantly reduces the workload for data scientists and democratizes ML for non-experts. Evolutionary algorithm (EA)-based automated ML (AutoML) systems, a prominent category of AutoML, often face inefficiencies due to the costly fitness evaluation of candidate ML pipelines. Although surrogate models have been employed to approximate the true performance of pipelines more quickly, a key challenge remains in effectively bridging the semantic gap between the heterogeneous features of datasets and pipelines. To address this issue, we propose ADELA, a novel accompanying surrogate-based optimization strategy that accelerates EA-based AutoML while retaining the performance of the resulting pipelines. ADELA operates in two phases: Offline, leveraging a high-quality curated pipeline corpus to meta-learn an accompanying surrogate model; and Online, selecting the accompanying pipeline and using the learned model to predict the performance of evaluation pipelines instead of executing them. The accompanying mechanism effectively mitigates the semantic gap between datasets and pipelines, enabling ADELA to reduce computation times by an average of 73.66% while retaining 98.78% of the final pipeline performance, as demonstrated in extensive experimental evaluations.

Practical applications often require forecasting the future states of short time series (STS) using multiple related long time series (LTS) as auxiliary data, a process known as few-shot prediction. The primary challenge, given the limited data on STS, is effectively capturing the pattern similarities between STS and LTS. Current methods, despite notable advancements, primarily focus on trans- ferring pattern characteristics from LTS to STS without explicitly addressing their similarities at various levels. To overcome this lim- itation, we propose a novel few-shot time series forecasting model called DANet. Drawing on the Retrieval-Augmented Generation (RAG) framework in large language models, DANet retrieves long and short sequences from LTS that closely resemble STS, thereby enhancing prediction accuracy while simultaneously reducing un- certainty through this retrieval process. First, we define two metrics to quantify pattern similarities between STS and LTS, addressing the issue of different representations of the same pattern due to variations in sequence length. Second, we propose a dual-attention mechanism which embeds the two similarities metrics to extract and integrate long and short sequences from LTS across variable and temporal levels for generating predictions. Our experiments across six scenarios show that DANet significantly outperforms six state-of-the-art (SOTA) methods.

Graph-based fraud detection has emerged as a pivotal tool in risk management, leveraging the power of graph neural networks to enhance node representations by aggregating information from neighboring nodes. However, this aggregation process can sometimes introduce noise by incorporating neighbors from different categories, potentially diluting the central node’s representation. To tackle this issue, we introduce an innovative Adversarial Confidence Reweighting (ACR) technique designed to allocate discriminative weights to samples automatically. This approach effectively minimizes the impact of noisy neighbors on the representation of nodes. By introducing controlled adversarial perturbations to the nodes being classified, we can assess the extent of representation dilution. Furthermore, we employ an adaptive node re-weighted learning objective, which dynamically adjusts node weights based on a confidence measure derived from prediction accuracy. Our experimental evaluations across three public datasets demonstrate that our adversarial strategy significantly surpasses the baseline model in terms of detection performance.

An effective auto-scaling framework is essential for microservices to ensure performance stability and resource efficiency under dynamic workloads. As revealed by many prior studies, the key to efficient auto-scaling lies in accurately learning performance patterns, i.e., the relationship between performance metrics and workloads in data-driven schemes. However, we notice that there are two significant challenges in characterizing performance patterns for large-scale microservices. Firstly, diverse microservices demonstrate varying sensitivities to heterogeneous machines, causing difficulty in quantifying the performance difference in a fixed manner. Secondly, frequent version upgrades of microservices result in uncertain changes in performance patterns, known as pattern drifts, leading to imprecise resource capacity estimation issues. To address these challenges, we propose Humas, a heterogeneity- and upgrade-aware auto-scaling framework for large-scale microservices. Firstly, Humas quantifies the difference in resource efficiency among heterogeneous machines for various microservices online and normalizes their resources in standard units. Additionally, Humas develops a least-squares density-difference (LSDD) based algorithm to identify pattern drifts caused by upgrades. Lastly, Humas generates capacity adjustment plans for microservices based on the latest performance patterns and predicted workloads. The experiment results conducted on 50 real microservices with over 11,000 containers demonstrate that Humas improves resource efficiency and performance stability by approximately 30.4% and 48.0%, respectively, compared to state-of-the-art approaches.

The popularity of open source has witnessed significant growth in the past two decades. The sustainability and success of open source projects rely heavily on the active community of contributors, who have the freedom to join or depart voluntarily. The churn of core developers can have detrimental effects on the continuous development and progress of these communities. Therefore, previous studies have aimed to identify the factors that influence developer churn and reveal its impact on open source communities. However, there is currently a lack of research on predicting whether a developer will leave or retain within a given timeframe. This article aims to bridge this gap by introducing a novel model called context-aware dynamic graph neural network (ConDyGNN), which is designed specifically for predicting developer churn in open source communities. By leveraging historical data, this model takes into consideration various factors such as developers’ activities, collaborative networks, and community context information. To capture both temporal dependencies and spatial correlations among different developers, temporal convolutional network (TCN) and graph convolutional network (GCN) modules are employed inConDyGNN. Furthermore, a context-aware module based on a multihead attention mechanism is proposed to effectively integrate macrolevel community information and enhance the predictive performance. The superiority of the proposed model is empirically demonstrated through extensive comparative experiments and an ablation study.

Map matching is crucial to facilitating location-based services, and recent advancements in map matching have demonstrated excellent performance with high-quality data. However, the use of low-precision devices often introduces high measurement noise, and the slow update rate of maps may result in errors in digital maps. Consequently, multiple types of noise significantly impact the performance of map matching algorithms. To tackle this issue, this paper presents a novel multi-noise perception framework, named MPF, aiming to enhance the performance and robustness of existing map matching algorithms. The main challenge lies in detecting anomalies during map matching, identifying the root causes, and devising appropriate solutions. Firstly, we propose a matching quality assessment (MQA) method that assesses abnormal variance in matching probability. Secondly, we introduce a multiple noise discrimination (MND) mechanism to effectively differentiate between measurement noise and map errors. Thirdly, we present a missing segment generation (MSG) scheme that dynamically fills in map gaps to prevent significant detours. To validate the effectiveness of MPF, we conduct experiments using real-world taxi trajectories from four cities, covering a total distance of 79,670.6 km. MPF is compare with seven online map matching algorithms and is used to optimize their performance. The experiments show that MPF outperforms the top baselines by 15.6%-26.9% and enhances their performance by 18.7%-38.2%.

The recent years have witnessed a surge in the development of traffic flow prediction methods, often deployed on cloud platforms to offer predictive services for entire transportation networks. However, the processes of training and executing a model for the entire traffic network are both time-consuming and computationally expensive. As a result, the utilization of edge servers for local sub-network prediction services has gained prominence. Nevertheless, training prediction models for numerous sub-networks within the extensive traffic network remains a time-intensive and computing resource-consuming task. To tackle this challenge, this article introduces the Pre-trained model REcommendation Framework for Edge computing enabled tRaffic flow prediction (PREFER). PREFER trains a set of traffic flow prediction models on selected sub-networks, then recommends optimal pre-trained models for edge servers. The recommendation is specifically based on performance prediction, integrating neural collaborative filtering and traffic flow characteristics. Experiments conducted on real datasets reveal that the pre-trained models recommended by PREFER perform close to the actual optimal ones and significantly outperform existing recommendation algorithms.

Event sequence anomaly detection has garnered considerable attention in research, encompassing applications such as identifying anomalies in system logs, anomalous transaction users, and so on. Yet, prevailing anomaly detection methods often rely solely on local data for training, potentially leading to imperfect detection performance. In this article, we introduce a personalized Federated anomaly detection framework for discrete event Sequences, named FeadSeq. Specifically, we propose a separate architecture for sequence reconstruction networks (SEPRE) which partitions the network into two parts: a shared part and a standalone part, better suited for federated learning schemes. In tandem, we propose a novel partial shared federated learning scheme that employs a mask strategy to alleviate communication overhead and produce personalized local models to address the statistical heterogeneity of data among clients. This scheme dictates that a subset of weights is communicated between clients and servers for collaborative training, while the remaining weights are trained exclusively locally. To evaluate the effectiveness of FeadSeq, we conduct extensive experiments on both system logs and business process event logs. The results affirm the superiority of FeadSeq over existing personalized federated learning algorithms, showcasing not only improved performance but also reduced communication overhead.

Multivariate time series forecasting is applied in many domains, such as finance, transportation, and industry. The main challenge of precise forecasting lies in accurately capturing latent dependencies. Recent studies develop various frameworks to reduce computational complexity or to enhance the learning of intricate relationships, while lacking interpretability and generality. In this article, we aim to elucidate the capture of dependencies as the recognition of patterns. We believe that patterns can be formally described from two aspects: the shapes of segments that frequently repeat and the corresponding forms of repetitions. Drawing upon this idea, we design a multivariate time series forecasting model named PRformer , 1 which incorporates a pattern-oriented attention mechanism and a pattern-based projector. The attention mechanism can perceive different forms of repetitions by embedded with various similarity evaluation metrics between segments, and filter out noise from segments to extract potential patterns with a statistical-driven weighting scheme. The pattern-based projector is employed to form the forecasting results by deriving the representative patterns from the set of potential ones. By incorporating explicit definitions of patterns, PRformer is interpretable and general to various time series scenarios. Experimental results on seven datasets demonstrate that PRformer outperforms six state-of-the-art models by about 10.7% in forecasting accuracy.

Data imputation is crucial in the analysis of incomplete time series, such as forecasting and classification, which involves learning dependencies among the observed values to infer missing ones. As there are no ground truths for missing values, the challenge of time series imputation lies in preventing the model from overfitting to spurious correlations. In this article, we believe that ensuring dependency consistency between observed and imputed values in a sequence is paramount for data imputation. Based on this idea, we propose a model called IR 2 -Net , 1 which combines an incomplete representation mechanism (IRM) with an iterative reconstruction framework (IRF) to establish a closed-loop learning-validation imputation paradigm. Firstly, IRM facilitates the representation of dependencies in incomplete sequences while preserving their distributions and semantics, effectively preventing the model from capturing spurious correlations. Secondly, IRF enables the model to reconstruct identical complete sequences separately based on imputed and observed values, ensuring that the dependencies of imputed values remain consistent with those of the observed ones. We conduct experiments on four datasets and compare IR 2 -Net with seven state-of-the-art imputation models. The experiment results show that IR 2 -Net outperforms all the baselines by 4.1%–23.4% in terms of accuracy. Moreover, IRF and IRM are two general modules that can be easily integrated into two existing models, significantly enhancing their performance by 18.3%–42.0%.

Effective management of business processes is crucial for organizational success. However, despite meticulous design and implementation, anomalies are inevitable and can result in inefficiencies, delays, or even significant financial losses. Numerous methods for detecting anomalies in business processes have been proposed recently. However, there is no comprehensive benchmark to evaluate these methods. Consequently, the relative merits of each method remain unclear due to differences in their experimental setup, choice of datasets and evaluation measures. In this paper, we present a systematic literature review and taxonomy of business process anomaly detection methods. Additionally, we select at least one method from each category, resulting in 16 methods that are cross-benchmarked against 32 synthetic logs and 19 real-life logs from different industry domains. Our analysis provides insights into the strengths and weaknesses of different anomaly detection methods. Ultimately, our findings can help researchers and practitioners in the field of process mining make informed decisions when selecting and applying anomaly detection methods to real-life business scenarios. Finally, some future directions are discussed in order to promote the evolution of business process anomaly detection.

Event matching is a key component in a large-scale content-based publish/subscribe system. The performance of most existing algorithms is easily affected by the subscription matching probability. In this paper, we propose a new data structure, named AWAW B+-Tree, which is based on the width of the predicates, to efficiently index the subscriptions. The most notable feature ofAW B+-Treeis its ability to combine the advantages of different matching methods, thus achieving high and robust performance in dynamic environments. First, we implement both a forward matching method (AFM) and a backward matching method (ABM) based onAW B+-Tree. Then, we introduce a hybrid matching method (AHM) that combines AFM and ABM. Moreover, we extendAW B+-Treein three aspects: approximate matching, string type matching, and fine-grained parallelization. We conducted extensive experiments to evaluate the performance of the proposed matching algorithms on synthetic and real-world datasets. The experiment results reveal that AHM achieves a reduction in matching time by up to 53.8% compared to the state-of-the-art method. Additionally, AHM exhibits improved performance robustness, with up to a 76.9% reduction in terms of the standard deviation of matching time. Particularly in dynamic scenarios, AHM is at least 2.3 times faster and 41.3% more stable than its counterparts. Furthermore, by implementing parallelization, the matching speed of 8 threads can be accelerated by 4.16 times compared to the single-thread matching speed.

Content-based publish/subscribe systems have the capability to achieve fine-grained data distribution, rapidly forwarding data from publishers to subscribers with specific requirements. The event matching algorithm is a fundamental component, quickly searching for subscriptions that match an event based on constraints defined by subscribers. As data scales continue to expand, there is a heightened demands for more efficient, robust, and versatile event matching algorithms. In this paper, we propose a novel data structure called the Dual-ended Interval Tree (DI-Tree). Firstly, given the splitting point, this data structure classifies intervals into three categories based on the joint distribution of their left and right endpoints in the attribute value domain. Furthermore, the DI-Tree utilizes blue and green nodes to store these three categories of intervals, resulting in enhanced indexing efficiency. Moreover, by utilizing the DITree to efficiently search matching and unmatching intervals, we develop innovative forward and backward event matching algorithms. Additionally, to enhance matching efficacy and reduce memory usage, we introduce key optimization techniques, focusing on improving node balance and bitset optimization. We conduct extensive experiments to evaluate the performance of DITree. When compared with five state-of-the-art event matching algorithms, the DI-Tree demonstrates an average reduction of up to 76.8 % in terms of matching time. This significant improvement highlights the effectiveness of our proposed strategies and the potential of DI-Tree in optimizing event matching performance.

It is time-consuming and knowledge-intensive for scientists to find practical workflows from the massive number of scientific workflow models. Currently, the retrieval approaches are mainly based on text matching between natural language queries and the descriptions of candidate workflows. Notably, the workflow structure also provides essential semantics, but the challenge lies in effectively matching these two pieces of heterogeneous information. To address this issue, we propose a Proxy-based Scientific workflow retrieval approach, ProSwats, which selects a workflow as the Proxy for each text query to bridge the gap between textual and structural semantics. ProSwats consists of two stages: workflow pre-selection based on text similarity and workflow ranking based on a matching degree prediction model. The textual and structural features are integrated by the proxy in this model, which is used to predict and rank the degree of semantic matching between the user query and candidate workflows. Crucially, ProSwats incorporates a confidence-aware learning mechanism to adapt to the varying reliability of proxies, enhancing generalizability. Extensive experimental results on two real-world datasets demonstrate that ProSwats outperforms state-of-the-art methods with statistical significance.

In business processes, anomalies are prevalent, arising from diverse factors, such as software malfunctions and operator errors. Detecting these anomalies is imperative, as it significantly influences not only the financial well-being of a business but also the dependability of event logs for subsequent analysis. However, existing deep business process anomaly detection approaches either lack effective temporal dependency modeling between events or encounter challenges associated with gradient vanishing. In this paper, we introduce COMB, which stands for an interConnected transfOrmers-based autoencoder for Multi-perspective Business process anomaly detection. Considering the interdependent nature of multi-perspectives, COMB leverages multiple parallel and interconnected transformers, facilitated by the inclusion of aggregation layers. These layers serve as integration points for information from various perspectives. Notably, considering how control flow influences other perspectives, COMB incorporates innovative mask adapters to enhance its detection performance. Furthermore, we propose a novel method for calculating anomaly scores, which effectively mitigates the influence of varying numbers of potential attribute values. Our extensive experimental evaluation encompasses both synthetic and real-life logs, and the results clearly demonstrate that COMB outperforms state-of-the-art methods in both trace-level and attribute-level anomaly detection.

Summary Finding similar scientific workflow modules that can substitute essential components of the privacy workflows from public repositories to create a personalized workflow is growing in popularity. This is a cost‐effective and error‐free strategy for the scientific community. Currently, module alignment approaches heavily depend on syntactic information such as modules' names and types. However, the contextual semantic information of modules, which encompasses inputs, outputs, and datalinks connecting modules, can also convey their functions. Unfortunately, this information is scarcely utilized during the module alignment process. In this work, we propose a module alignment method based on neighbor information for scientific workflow (MANSOR). Specifically, we present a rule‐based attribute similarity computation approach for calculating initial module‐pair similarity in two workflows. The relation similarity is then employed to iteratively fine‐tune the matching degree with uncertainty of module pairs by considering the contextual semantics until reaching a steady state. After pairwise comparison of workflows, the module alignment results are determined based on their module‐pair similarity. Experimental results on the human‐curated corpus of ratings indicate that MANSOR outperforms the existing state‐of‐the‐art approaches with statistical significance.

The ability to detect anomalies in business processes is crucial for achieving success in business operations. While unsupervised anomaly detection approaches have gained popularity in recent years due to their label-free nature, in some cases, a limited number of labelled anomalies can be provided and using them can improve the performance of anomaly detection. To address this issue, we propose a novel framework for anomaly detection that uses a pre-trained autoencoder to extract feature representations of traces. An anomaly score generator based on a multi-layer perceptron is utilized to evaluate the extracted features. The entire framework is trained using a joint loss that ensures the generated anomaly scores satisfy a specific distribution without compromising the autoencoder's ability to reconstruct normal traces. The feature encoder is fine-tuned to provide insights into the cause of anomalies. Additionally, we design a novel technique for calculating anomaly scores to mitigate the effects of varying numbers of potential attribute values. We conduct extensive experiments on both synthetic and real-life logs, and our results demonstrate that our proposed method, WAKE, outperforms state-of-the-art unsupervised deep business process anomaly detection methods by a significant margin. Additionally, it outperforms other weakly supervised anomaly detection methods as well

For software-defined networking (SDN), multi-field packet classification plays a key role in the processing of flows, mainly involving fast packet classification and dynamic rule updates. Due to the increasing complexity and size of rulesets, it is becoming more difficult to design a packet classification algorithm which achieves fast lookup and update. In this paper, we propose a novel structure, PT-Tree, for packet classification with high overall performance. PT-Tree cascades the prefixes of multiple discriminatory bytes to achieve efficient partitioning of the ruleset, thereby reducing the search space and ensuring the performance of both lookup and update. Meanwhile, a multi-granularity priority-aware pruning mechanism (MPPM) based on PT-Tree filters out most of the candidate subsets, which further improves the lookup speed. In addition, we propose an auxiliary tree-based optimization method (ATOM) to cope with severely overlapping rules in the search space. Therefore, PT-Tree can better handle the case where the rules in certain fields are skewed. We conduct comprehensive experiments to evaluate the performance of PT-Tree. The results show that compared with the state-of-the-art, the lookup time of PT-Tree is reduced by at least 49.95% on average. Moreover, PT-Tree is also at least 7.13x and 33x faster than the baselines in terms of the update and construction speed on average, respectively. Meanwhile, the performance stability of PT-Tree on multiple rulesets improves by up to 13.68 times.

Packet classification, as a crucial function of networks, has been extensively investigated. In recent years, the rapid advancement of software-defined networking (SDN) has introduced new demands for packet classification, particularly in supporting dynamic rule updates and fast lookup. This paper presents a novel structure called DBTable for efficient packet classification to achieve high overall performance. DBTable integrates the strengths of conventional packet classification methods and neural network concepts. Within DBTable, a straightforward indexing scheme is proposed to eliminate rule replication, thereby ensuring high update performance. Additionally, we propose an iterative method for generating a discriminative bitset (DBS) to evenly partition rules. By utilizing the DBS, rules can be efficiently mapped in a hash table, thus achieving exceptional lookup performance. Moreover, DBTable incorporates a hybrid structure to further optimize the worst-case lookup performance, primarily caused by data skewness. The experiment results on 12 256k rulesets show that, compared to seven state-of-the-art schemes, DBTable achieves an overall lookup speed improvement ranging from 1.53x to 7.29x, while maintaining the fastest update speed.

Content-based publish/subscribe systems are widely used in many fields.Event matching is the core component to achieve fine-grained content-based data distribution.Many efficient algorithms have been proposed to improve event matching performance.However, in large-scale content-based publish/subscribe systems, event matching is still the performance bottleneck of the entire system due to the need to perform a lot of operations, such as additions, comparisons and bitmarkings.In this paper, we explore to convert various nonlogical operations into efficient logical ones, and propose a bitsetbased optimization paradigm (BOP) for matching algorithms.On the one hand, BOP can eliminate expensive operations in the matching process, greatly improving matching performance.On the other hand, BOP can stabilize the performance of matching algorithms, ensuring the quality of service of data distribution.We apply BOP to optimize two existing matching algorithms, namely TAMA and REIN.The experimental results show that BOP shortens the matching time of TAMA and REIN by more than 60%.In addition, the performance of optimized versions is more stable than the original matching algorithms.

Accurately forecasting workloads in terms of throughput that is quantified as queries per second (QPS) is essential for microservices to elastically adjust their resource allocations. However, long-term QPS prediction is challenging in two aspects: 1) generality across various services with different temporal patterns, 2) characterization of intricate QPS sequences which are entangled by multiple components. In this paper, we propose a knowledge auto-embedding Informer network (KAE-Informer) for forecasting the long-term QPS sequences of microservices. By analyzing a large number of microservice traces, we discover that there are two main decomposable and predictable components in QPS sequences, namely global trend & dominant periodicity (TP) and low-frequency residual patterns with long-range dependencies. These two components are important for accurately forecasting long-term QPS. First, KAE-Informer embeds the knowledge of TP components through mathematical modeling. Second, KAE-Informer designs a convolution ProbSparse self-attention mechanism and a multi-layer event discrimination scheme to extract and embed the knowledge of local context awareness and event regression effect implied in residual components, respectively. We conduct experiments based on three real datasets including a QPS dataset collected from 40 microservices. The experiment results show that KAE-Informer achieves a reduction of MAPE, MAE and RMSE by about 16.6%, 17.6% and 23.1% respectively, compared to the state-of-the-art models.

Summary When large‐scale content‐based publish/subscribe systems face dynamic workloads, it is challenging to stabilize event delivery latency. In this article, we propose a latency‐aware parallelism framework (Lap) to address this challenge. Specifically, we propose a lightweight parallelism method called PhSIH for event matching algorithms. In addition, we design a reactive parallelism degree adjustment (RPDA) mechanism in the backpressure way to determine the parallelism degree. We implement Lap in Apache Kafka and evaluate the parallelism effect of PhSIH and the adaptability of RPDA on trace data. The experiment results demonstrate that PhSIH achieves linear speedup on three existing algorithms and RPDA possesses a desirable adaptability to the dynamic workloads.

Online map matching is essential for some location-based services, such as car navigation. However, due to GPS measurement errors and/or the lack of sufficient information, the performance of most existing online algorithms will degrade in the increasingly complex traffic environment. In this paper, we propose an adaptive online map matching algorithm called AMM. The basic idea is that AMM should be able to calibrate GPS observation data for various measurement errors and under complex urban conditions. First, we establish a collaborative evaluation model between GPS points and candidate points to effectively filter low-quality GPS measurement points, dynamically set the weights of different features and comprehensively select the best candidate points. Second, we propose a retrospective correction mechanism to correct the previous matching results when more information is available, which will help improve the accuracy of future GPS points. Furthermore, we define parameter self-tuning rules for AMM to enhance its portability by avoiding time-consuming parameter tuning steps. We conduct extensive experiments to evaluate the performance of AMM on real vehicle trajectory datasets. The experiment results show that AMM outperforms its counterparts by up to 32% in terms of accuracy and its performance in different traffic conditions is more stable.

A session-based recommendation system (SRS) tries to predict the next possible choice of anonymous users. In recent years, graph neural network (GNN) models have been successfully applied to SRSs and have achieved great success. Using GNN models in SRSs, each session graph is processed successively to obtain the embedding of the node (i.e, each action on an item), which is then imported into the prediction module to generate recommendation results. However, solely depending on the session graph to obtain the node embeddings is not sufficient because each session only involves a few items. Therefore, neighbor sessions have been used to extend the session graph to learn more informative node representations. In this paper, we introduce a S ession-based recommendation MO del based on N eighbor sessions with similar probabilistic int E ntions(SMONE). SMONE models the intentions behind sessions in a probabilistic way and retrieves the neighbor sessions with similar intentions. After the neighbor sessions are found, the target session and its neighbor sessions are modeled as a hypyergraph to learn the contextualized embeddings, which are combined with item embeddings through GNN to produce the final item recommendations. Experiments on real-world datasets prove the effectiveness and superiority of SMONE.

Predictive monitoring of business processes aims at forecasting the future information of a business process and has gained increasing attention in recent years. With the development of cloud computing, prediction models, including remaining time prediction models, can be provided as cloud services. Reducing the number of invocation times of a remaining time prediction service is necessary since a large quantity of business process instances may be initiated and monitored every day. However, most of the current research focuses on designing new algorithms to improve prediction accuracy but there is no approach available to decide when to invoke a prediction service for each business process instance. In this article, we propose a deep reinforcement learning based strategy that can learn the policies of selecting prediction points for the remaining time prediction. Specifically, the learned policies can dynamically decide the next prediction point at which the remaining time prediction service will be invoked for a business process instance. We performed extensive experiments on five real-world datasets. The experiment results show this strategy can reduce the number of prediction points significantly and still maintain the high prediction accuracy.

Finding scientific workflow models that can be reused or repurposed from public repositories is becoming popular in the scientific community. Currently, the retrieval approaches for workflow models are mainly based on text matching between queries and the descriptions of models. However, the structure information of these models, which includes inputs, outputs, data processing modules, and datalinks connecting modules, expresses their functions in a more detailed manner, yet this information is not used in the model retrieval process. Therefore, we propose a two-stage framework for Scientific WOrkflow Retrieval by learning Textual and Structural semantics (SWORTS). The framework comprises a workflow pre-selection step and a workflow ranking step. Specifically, we use text similarity approaches to quickly identify candidate workflow models in the first step. Then, a hierarchy-based matching degree prediction model, which takes both textual and structural features into account, is trained to predict and rank the degree of semantic matching between requirement specifications and candidate workflows. The experiment results on real-world datasets demonstrate that SWORTS achieves the best performance with relatively balanced effectiveness and efficiency among state-of-the-art methods.

In process-aware information systems (PAISs), anomalies are ubiquitous, having a number of different underlying causes, such as software malfunctions or operator errors. The presence of anomalies not only has an enormous impact on the economic well-being of the business, but also interferes with our ability to mine useful information from event logs. In this article, we propose GRASPED, aGRU-AE Network based multi-perSpective business ProcEss anomalyDetection model. GRASPED can detect anomalies not only from the control flow but also from the data perspective of a business process. GRASPED is based on an autoencoder (AE) with gated recurrent units (GRU) which is trained in an unsupervised fashion (i.e., does not require any labeling of the data). In addition, GRASPED introduces the teacher forcing method as well as the attention mechanism to improve its detection performance. GRASPED does not require training on a clean log (i.e., it can be trained and perform anomaly detection directly on logs containing anomalies). We conduct extensive experiments on synthetic logs as well as real-life logs. The experiment results show that GRASPED outperforms the state-of-the-art methods for both trace-level and attribute-level anomaly detection.

Multi-task learning (MTL) aims at solving multiple related tasks simultaneously and has experienced rapid growth in recent years. However, MTL models often suffer from performance degeneration with negative transfer due to learning several tasks simultaneously. Some related work attributed the source of the problem is the conflicting gradients. In this case, it is needed to select useful gradient updates for all tasks carefully. To this end, we propose a novel optimization approach for MTL, named GDOD, which manipulates gradients of each task using an orthogonal basis decomposed from the span of all task gradients. GDOD decomposes gradients into task-shared and task-conflict components explicitly and adopts a general update rule for avoiding interference across all task gradients. This allows guiding the update directions depending on the task-shared components. Moreover, we prove the convergence of GDOD theoretically under both convex and non-convex assumptions. Experiment results on several multi-task datasets not only demonstrate the significant improvement of GDOD performed to existing MTL models but also prove that our algorithm outperforms state-of-the-art optimization methods in terms of AUC and Logloss metrics.

Multi-task learning (MTL) has experienced rapid growth in recent years. A typical way of conducting MTL with deep neural networks (DNNs) is either establishing a sort of global feature sharing mechanism across all tasks or assigning each task an individual set of parameters with cross-connections. However, these existing approaches leverage DNNs only to share features of a certain order. Several modelsdemonstrated that explicitly modeling feature sharing with both low-order and high-order features can boost performance. To this end, we propose a model-agnostic sparse routing architecture called MASR, which emphasizes arbitrary order feature sharing for multi-task learning. It is able to choose specific orders of features to route for a given task through learnable latent variables. Moreover, MASR is model-agnostic and can be combined with existing MTL models to share features of both low-order and high-order. Extensive experimental results on several real-world datasets not only confirm the significant improvement of MASR performed to existing MTL models but also outperform existing hybrid architectures in terms of AUC metric.

When incoming events arrive too quickly, edge nodes cannot process all events since they only have limited resources. Therefore, we propose a CLoud-edge dynamic cOllaborative Strategy for complex Event Detection (CLOSED). We consider the case where a composite event may involve primitive events from more than one edge node and model the composite event detection task allocation as an optimization problem to minimize the communication costs for both event detection and event notification among cloud centers and edge nodes. We propose the MINimized Event Transmission time-based detection node selection algorithm (MINET) to solve this problem. Moreover, we propose a Reward-based composite event detection Task Migration Algorithm (RTMA) to migrate event detection tasks when edge nodes are overloaded. By evaluating CLOSED on a simulated dataset and different workload patterns , we prove that this strategy can significantly improve event detection efficiency.

Data analysis reveals that passengers can only endure a few minutes before taking a taxi. However, most existing route recommendation systems are not adequate to satisfy impatient customers due to two shortcomings: inaccurate demand forecast and the lack of an efficient supply-demand balance mechanism. In this paper, we propose a recommendation system called @ Me, which aims to dispatch vacant taxis to the vicinity of potential customers at the right time. To achieve minute-level demand forecasting, we ensemble a contextualized spatial-temporal network (CSTN) with an LSTM network to optimize prediction accuracy. In addition, we characterize the attractiveness of the road grid to vacant taxis as a force model, on which a taxi scheduling algorithm is proposed to dynamically balance supply and demand. Extensive experiments on real datasets clearly indicate that our method is superior to the selected baselines. Vacant taxis that follow the routes suggested by @ME can catch more impatient customers in a shorter cruising time. The 7 -day experimental results on the Manhattan dataset show that @Me can carry an additional 48,332 passengers and increase drivers' revenue by $570,317.

Optical camera communication (OCC) has attracted increasing attention recently thanks to the wide usage of LED and high-resolution cameras. The lens-image sensor structure enables the camera distinguish light from various source, which is ideal for spatial MIMO. Hence, OCC can be applied to several emerging application scenarios, such as vehicle and drone communications. However, distance is a major bottleneck for OCC system, because the increase in distance makes it difficult for the camera to distinguish adjacent LEDs, which we call LED spatial mixing.In this paper, we propose a novel hierarchical coding scheme named as OnionCode to support dynamic range of channel capacity in one-to-many OCC scenario. OnionCode adopts a multi-priority receiving scheme, i.e., the receivers can dynamically discard the low-priority bit streams according to the measured channel capacity. OnionCode achieves this based on a key insight that, the luminance level of a mix-LED is distinguishable. We prototype a LED-based OCC system to evaluate the efficacy of OnionCode and the results show that OnionCode achieves a higher coding efficiency and overall throughput compared with the existing hierarchical coding.

Content-based publish/subscribe systems are an effective paradigm for implementing on-demand event distribution.Each event needs to be matched against subscriptions to identify the target subscribers.To improve the matching performance, many novel data structures have been proposed.However, the predicates included in subscriptions are handled the same way in most existing data structures, which is not efficient given the matching probability of predicates.In this paper, we propose a parallel ensemble matching framework called PEM, which uses multiple algorithms with complementary behavior on predicate matching probabilities.To achieve the performance balance of parallel matching, we design an elastic subscription classification method.We implement a prototype of PEM based on two existing algorithms.The experiment results show that PEM improves the matching performance by 43%.

The content-based publish/subscribe system is an effective paradigm for implementing on-demand event distribution. Each event needs to be matched against subscriptions to identify the target subscribers. To improve the matching performance, many novel data structures have been proposed. However, the predicates contained in subscriptions are handled the same way in most existing data structures, without considering their differences in matching probability. In this paper, we propose the concept of parallel ensemble matching (PEM) based on subscription partitioning. The basic idea is that we have the right algorithm handling the right subscriptions at the right time. First of all, we design a PEM framework by classifying subscriptions according to their matching probabilities and use the proper algorithms to process each subscription category. Furthermore, to deal with high-dimensional subscriptions, we propose a fine-grained PEM (fgPEM) that exploits matching algorithms with complementary behaviors by partitioning subscriptions into sub-subscriptions. We implement the prototype of PEM and fgPEM based on two existing algorithms. The experiment results show that PEM improves the matching performance by 43%. On the basis of PEM, fgPEM further improves the performance by 31%.

Researchers have proposed many novel methods to detect abnormal taxi trajectories. However, most of the existing methods usually adopt a counting-based strategy, which may cause high false positives due to imprecisely identifying diverse trajectories as anomalies and therefore, they need the support of large-scale historical trajectories to work properly. To improve detection precision and efficiency, in this article, we propose STR, an online abnormal taxi trajectory detection method based on spatio-temporal relations. The basic principle behind STR is that given the displacement from the source point to a testing point, if the driving time and driving distance are not within the normal ranges, the point is identified as anomalous. To learn the two normal ranges for driving time and driving distance, STR defines two spatio-temporal models which characterize the relationship between displacement and driving distance/driving time. To improve detection efficiency, STR reduces the number of models that need to be learned by making full use of the similarity of transportation modes in different time periods and neighboring areas. The effectiveness and performance of STR are evaluated on real-world taxi trajectories. The experiment results show that compared with counting-based methods, STR achieves greater precision by reducing false positives. Furthermore, STR is more efficient than its counterparts and is suitable for online detection.

With the development of multiaccess edge computing (MEC) technology at the network edge, efficient resources allocation and offloading between the resource-constrained edge clouds to maintain load balancing become a looming problem recently. However, most existing researches aimed at optimizing the allocation of computing resources are based on simulation and apply only for some typical Internet-of-Things (IoT) applications on mobile devices (i.e., they all lack of practicality and generality). In this article, we present a novel edge computing platform (CampEdge) with 36 edge nodes based on the wireless access point (AP) for adaptive resources allocation and computation offloading in a complicated dynamic campus environment. We first collected and sufficiently analyzed a large real-world WiFi data set, covering more than 8500 wireless APs and serves 44 000 active end users within an area of 3.1 km2over three months. A multiclass classification algorithm based on the random forest was then used with this data to accurately predict the state of resources usage at each edge node (i.e., in a busy state or normal state). To reasonably offload and transmit end-users' computational tasks to these edge nodes and optimize the total latency cost among the whole offloading process, we then illustrate a distributed computation offloading optimization strategy to formulate this complicated problem as a multiobjective latency optimization problem based on alternating direction method of multipliers (ADMM). Furthermore, we briefly discuss the convergence of our system. In experiments, CampEdge was shown to decrease user latency by up to 30%, compared to state-of-the-art methods. The proposed strategy was also shown to adaptively conform to a variety of compute-intensive and time-sensitive IoT applications for end users.

For large-scale spatiotemporal-aware publish/subscribe systems, it is critical to design an efficient forwarding engine to achieve fast matching and maintenance of events and subscriptions. For this goal, we propose a novel data structure called MO-Tree to index both subscriptions and events in a unified way. The design philosophy behind MO-Tree is to keep the data structure concise, which manifests in three aspects: limiting the height of MO-Tree, trading space for time, and avoiding node merging and splitting. The difficulty in designing MO-Tree is how to efficiently index width-variable intervals. We present a multi-level cell-overlapping partition scheme and build a theoretical model to optimize the cell width in each level. To evaluate the performance of MO-Tree, a series of experiments is conducted on real-world trace datasets. The experiment results show MO-Tree significantly outperforms the state-of-the-art in terms of matching speed and maintenance cost.

With the development of cloud computing and the advent of the Web 2.0 era, composing a set of Web services as a service process is becoming a common practice to provide more functional services. However, a service process involves multiple service invocations over the network, which incurs a huge time cost and could become a bottleneck to performance. To accelerate its execution, we propose an engine-side cache-block-based service process caching strategy (CBPCS). It is based on, and derives its advantages from, three key ideas. First, the invocation of Web services embodies semantics, which enables the application of semantic-based caching. Second, cache blocks are identified from a service process, and each block is equipped with a separate cache so that the time overhead of service invocation and caching can be minimized. Third, a replacement strategy is introduced taking into account time and space factors to manage the space allocation for a process with multiple caches. The algorithms and methods used in CBPCS are introduced in detail. Moreover, how CBPCS can be applied to multiple service process models is also investigated. Finally, CBPCS is validated via comparison experiments, which shows the considerable improvements of CBPCS over other strategies.

Multi-access edge computing (MEC) plays an important role in taking the cloud computing resources closer to the end-user at the network edge. The main challenge in such systems is to build models to efficiently predict and offload computational resources to maintain load balancing among edge clouds. In this paper, we present an Efficient Resources Allocation and Computation Offloading (ERACO) model with 36 edge nodes based on the large-scale wireless access point (AP) in a campus environment. We first collected and sufficiently analyzed this large WiFi dataset, covering more than 8,500 wireless APs and serves 44,000 active end-users within an area of 3.1 km2over three months. A multi-class classification algorithm was then used with this data to predict the usage of computing resources at each edge node. To reasonably offload these resources and decrease the latency between each edge node, we then propose the optimization strategy to formulate the complicated offloading problem as a multi-objective latency optimization problem using the alternating direction method of multipliers (ADMM). Experimental results show that the models decrease user latency by up to 30%, compared to state-of-the-art methods for a variety of applications.

Although many matching algorithms have been proposed to improve the matching efficiency of the content-based publish/subscribe system, existing work seldom consider the real-time of event dissemination from the perspective of event matching. On the basis of two existing matching algorithms, in this paper, we propose a subscription-classifying and structure-layering (SCSL) optimization method for matching algorithms, aiming to improve real-time by shortening the determining time of matching subscriptions. The basic idea of SCSL is that subscriptions with high matching probabilities should be processed first in the process of event matching and their storage positions in the data structure should be adjusted in line with changing probabilities. One challenge of SCSL is the trade-off that needs to be made between the gains of improving real-time performance by identifying matching subscriptions earlier and the cost of increasing matching time due to subscription classification and adjustment. We design a concise scheme to classify subscriptions, establish a lightweight adjustment mechanism to deal with dynamics and propose an efficient greedy algorithm to compute the adjustment solution, which alleviates the impact of SCSL on matching performance. The experiment results show that the 95thpercentile of the determining time of matching subscriptions is improved by about 70%. Furthermore, we integrate SCSL into Apache Kafka to augment it as a content-based publish/subscribe system and test the effect of SCSL based on real-world stock trace data, which witnesses about 40% improvement on the average event transfer latency and confirms that SCSL can effectively improve the real-time performance of content-based publish/subscribe systems.

Financial fraud detection is an important topic in business intelligence, and neural networks have been effectively applied to construct detection models for this problem. Nevertheless, the timeliness of fitted models degrades over time when they are deployed in online detection systems. To maintain model performance when the labels of new transactions are not available, we propose an incremental virtual learning (IVL) method to update neural networks continually. The basic idea of IVL is to make the output distribution of neural networks smoother in the adversarial direction by unlabeled data. IVL uses local distribution smoothness (LDS) as the loss function at the unsupervised incremental learning stage. When updating a neural network by IVL, unlabeled data can be fed into the neural network periodically, so IVL can be implemented in a real online system without violating time constraints. To evaluate the effectiveness of IVL, experiments are conducted on a real dataset, demonstrating that neural networks augmented by IVL (IVL-NN) perform better than unoptimized neural networks in term of area under the curve (AUC). Furthermore, the standard deviation of the daily AUCs improves up to 19.09%, which indicates that IVL improves the stability of neural networks.

Teamwork is very important to software development. There are many studies focusing on different aspects of teamwork in open source software projects, but neglecting the fact that most teams of open source projects are temporary and dependent on the context of one specific project. Whether the collaboration of such teams can extend to different projects is highly doubted. In contrast, we are interested in long-lasting socially connected teams, whose members have steady social connections and have collaborated with each other on multiple projects. Therefore, we mine Cross-Repository Socially Connected (CRSC) teams on GitHub, the largest open-source project hosting platform. Community detection methods are used to mine CRSC teams from the developer network and more than 20,000 CRSC teams are discovered on GitHub. The productivity of such teams and how the hosting repository may influence them are studied. Their preferences for repositories are investigated. Moreover, we study the structures of these teams using complex network analysis methods. Our results indicate that CRSC teams are stable, highly productive and mature. Therefore, open-source project owners and recruiters can pay more attention to such teams.

Oceans are important for scientific research and also for global economic and military security. Usually, wireless ad hoc networks are chosen to transform real-time data collected by ocean monitoring sensors (nodes). Due to the random motion of waves or the random direction of the wind, nodes in the network might become detached from the coverage of the network. In this case, the detached nodes can either send the collected data directly to the base station at the cost of consuming more energy or wait for a period of time to rejoin the network with the price of sacrificing the real time of the collected data. In this paper, we model the optimal waiting time for detached nodes before directly sending the data in the dynamic environment of ocean monitoring. For this purpose, we need to address two problems. The first is how to calculate the rate of coverage with a different number and different broadcast radii of nodes. The second is when a node detaches from the coverage of the network, how much time will it need to wait before it rejoins the network. We first establish the motion model of nodes, which is the basis to deduce the probability distribution of a certain time when the detached node rejoins the network. Based on the probability distribution, the waiting time of the detached nodes can be optimally determined, aiming to achieve a good balance between energy consumption and data timeliness. Finally, a series of simulations is conducted to validate the effectiveness of our proposed method.

Nowadays, the container-based virtualization technologies have become very popular due to lightweight nature, scalability, flexibility and others. Kubernetes is one of the most popular container cluster management systems, which enables users to deploy applications on the container easily, so more and more web applications are deployed in a Kubernetes clusters. However, a Kubernetes cluster is generally designed to handle the peak of workloads, so that most of resources are idle in usual time, which results in an huge waste of resource. Hence, it is necessary to design a system to improve the cluster resource utilization and promise Quality of Service (QoS) in a Kubernetes cluster. In this paper, we propose a generic system to dynamically adjust the scale of a Kubernetes cluster, which is able to reduce the waste of resource on the premise of QoS guarantee. The proposed system contains four modules: monitor module, QoS module, scaling module, and executing module. First, the monitor module uses two open-source tools, Heapster and InfluxDB, to monitor and store real-time status of a Kubernetes cluster. Then, to guarantee QoS in the Kubernetes cluster, the QoS module presents a method to automatically decide a threshold of CPU utilization that is able to meet requirements of a specific application. Next, the scaling module provides a cluster scaling algorithm to get an ideal number of nodes in the Kubernetes cluster, which is used to allocate resources in a cluster-level allocation. Finally, according to the ideal number of nodes, the executing module adjusts the scale of the Kubernetes cluster to carry out the application. Extensive experiments in real environments show that, on average, the proposed system improves CPU utilization of a Kubernetes cluster by 28.99%.

The matching algorithm is a critical component of the content-based publish/subscribe system, whose performance has direct effects on the QoS of the whole system. Aiming to improve and stabilize the matching performance, we propose a lightweight parallelization method called PhSIH on the basis of three existing algorithms. PhSIH fulfills Parallelization by horizontally Segmenting the Indexing Hierarchy of data structures to support multiple threads performing matching tasks in parallel on a common data structure. PhSIH can adaptively adjust the degree of parallelism according to the changing workloads in order to meet the performance requirement. The main work of PhSIH concerns dynamically adjusting the degree of parallelism and computing a task allocation solution for parallel threads. PhSIH is implemented in Apache Kafka to augment it as a content-based publish/subscribe system, which makes Kafka suitable for real-time fine-grained event dissemination scenarios, such as stock ticks. To evaluate the parallelization effect and adaptability of PhSIH, a series of experiments are conducted based on synthetic and real-world data. The experiment results demonstrate that PhSIH achieves a good parallelization effect on the three existing algorithms and possesses a desirable adaptability that stabilizes the performance of the matching algorithms.

With the extensive applications of Web APIs, to monitor and analyze personal Web API invocation behaviors is becoming more and more important. However, due to the fact that the users' Web API invocation behaviors are highly heterogeneous, it is extremely challenging to provide a unified framework. In the paper, we introduce a predictive Web API Invocation Behavior Monitoring System (AIMS). AIMS automatically analyzes the predictability of users' invocation behaviors and for the user whose behaviors can be predicted, an adaptive strategy is applied to generate the predictions in an adaptive way. In addition, a context-aware K-nearest neighbor classifier is applied to detect the anomaly of user's invocation behaviors. Experiments on both a real-world dataset and synthetic dataset show AIMS is efficient in analyzing personal API invocation behaviors.

Dredging is the removal of sediment from the bottom of lakes, rivers, harbors, and other water bodies. It is a routine necessity in waterways around the world because the natural process of sand and silt washing downstream results in the sediment gradually filling channels and harbors. However, during the dredging operation, some dredgers do not transport the sediment to the designated area as expected but throw it near the waterway meaning sediment may return to the waterway in a short period. This paper proposes an online dredging operation anomaly detection (DOAD) method to detect this kind of irregular behavior during the dredging operation based on automatic identification system (AIS) data. First, we establish a feature system to extract behavior features from AIS data. Furthermore, we jointly utilize t-distributed stochastic neighbor embedding (t-SNE) with neural networks and a Gaussian mixture model (GMM) to train a detection model in a semi-supervised way. Through the trained model, irregular behaviors can be efficiently detected in real time during the dredging operation. The effectiveness of DOAD is evaluated according to a series of experiments. To the best of our knowledge of the published literature, this work is the first to introduce the application of AIS data to detect irregular behaviors during dredging operations.

Deep learning is popular in many areas, but users must manually specify the resource configuration when submitting deep learning training jobs, usually over-provisioning resources. This kind of unreasonable resource configuration method results in slow training and low resource utilization. Therefore, it would be more convenient and efficient if users only need to specify the quality of service (QoS) for their jobs, and then the resources will be autoconfigured to meet the QoS. To satisfy this demand, we present a QoS-oriented scheduling and autoscaling framework that schedules and autoscales deep learning training jobs in the Kubernetes cluster. This paper focuses on the most important QoS requirement for deep learning training jobs: deadline. The goal of the framework is to guarantee that as many jobs as possible can be accomplished before their specified deadlines. To reach this goal, the framework schedules deep learning jobs by implementing a heuristic scheduling policy based on resource status and job deadline, and autoscales resource configuration by exploiting a characteristic of deep learning jobs: the predictability of training time. This predictability is used to predict whether a job can be accomplished before its deadline and estimate appropriate resource configuration if necessary. We implemented the framework by modifying the default scheduler of Kubernetes and conducted experiments to evaluate its performance. The experiment results show that our scheduling policy can improve the completion rate by 26% when the cluster resources are insufficient, and our autoscaling policy can improve the completion rate to 100% when the cluster resources are sufficient. We also show that the framework improves the utilization of allocated CPUs to 100%. Our proposed framework points to a new way of submitting and managing deep learning training jobs in the cluster.

When facing fluctuating workloads, can the performance of matching algorithms in a content-based publish/subscribe system be adjusted to adapt to the workloads? In this paper, we explore the idea of endowing matching algorithms with adaptability. The prerequisite for adaptability is to enable the matching algorithm to possess the ability to dynamically and quantitatively adjust its performance. We propose PSAM, a Predicate-Skipping Adjustment Mechanism that realizes dynamic performance adjustment by smoothly switching between exact matching and approximate matching, following the strategy of trading off matching precision in favor of matching speed. The PSAM mechanism is integrated into an existing matching algorithm, resulting in a performance-adjustable matching algorithm called Ada-Rein. To collaborate with Ada-Rein, we design PADA, a Performance Adjustment Decision Algorithm that is able to make proper performance adjustment plans in the presence of fluctuating workloads. The effectiveness of Ada-Rein and PADA is evaluated through a series of experiments based on both synthetic data and real-world stock traces. Experiment results show that adjusting the performance of Ada-Rein at the price of a small false positive rate, less than 0.1%, can shorten event latency by almost 2.1 times, which well demonstrates the feasibility of our exploratory idea.

Multipath profiling is to characterize multipath components of wireless channels, which can be done using Channel State Information (CSI) from WiFi devices. To do so with satisfactory accuracy, recent studies rely on either a large number of receiving antennas or large bandwidth. However, it is difficult for commodity WiFi devices to meet these requirements. In this paper, we propose a scheme, WiZoom, that can perform accurate multipath profiling using single-band CSI from commodity WiFi devices. In order to achieve accurate multipath profiling with limited bandwidth, WiZoom first incorporates the MUltiple SIgnal Classification (MUSIC) algorithm with CSI to estimate ToAs of multipath components, and then combines multiple antennas to improve the resolution of ToA estimation. WiZoom further estimates attenuations and phase shifts for multipath components using the ToAs. So far, multipath components are fully characterized, and all these estimated parameters form the multipath profile. We evaluate the performance of WiZoom using commodity WiFi devices in real environment, and results show that WiZoom achieves high accuracy in multipath profiling.

Summary Nowadays, with the development of cloud computing technology, an increasing number of enterprises or organizations have migrated applications to the cloud environment. Because of privacy concerns, a company may store sensitive data on a local server or a private cloud. As a result, the data analytics tasks have to be performed only in the local environment even when the analysis results will be shared to the outside. In this paper, we design a decentralized workflow system to speed up the execution of a Privacy‐Awareness Data Analytics application (PADA). Specifically, a workflow system with actor layer and engine layer for distributed execution of PADA proposed. The actors are responsible for performing actual data analysis tasks on the data side while the engines are deployed into different regions of the public cloud and communicate with actors to execute the whole workflow. Since engines in different regions have different communication latencies when invoking an actor, the optimization engine deployment scheme should be found to minimize the total execution time of a PADA. A path search‐based heuristic algorithm is designed to select suitable cloud regions to place the engines in the public cloud. The experimental results indicate that the proposed algorithm is effective in reducing application makespan.

With the rapid development of cloud computing, various types of cloud services are available in the marketplace. However, it remains a significant challenge for cloud users to find suitable services for two major reasons: (1) Providers are unable to offer services in complete accordance with their declared Service Level Agreements, and (2) it is difficult for customers to describe their requirements accurately. To help users select cloud services efficiently, this article presents a Trust enabled Self-Learning Agent Model for service Matching (TSLAM). TSLAM is a multi-agent-based three-layered cloud service market model, in which different categories of agents represent the corresponding cloud entities to perform market behaviors. The unique feature of brokers is that they are not only the service recommenders but also the participants of market competition. We equip brokers with a learning module enabling them to capture implicit service demands and find user preferences. Moreover, a distributed and lightweight trust model is designed to help cloud entities make service decisions. Extensive experiments prove that TSLAM is able to optimize the cloud service matching process and compared to the state-of-the-art studies, TSLAM improves user satisfaction and the transaction success rate by at least 10%.

A distributed cluster provides a common computation platform to efficiently run different types of jobs, such as real-time and batch jobs that have different QoS requirements. One challenge of a distributed cluster is to find a scheduling strategy that can reduce the overtime ratio of real-time jobs as much as possible while also improving the completion time of batch jobs, aiming to promote QoS even though the resources in the cluster are stretched. To address this issue, many efficient scheduling methods have been proposed. However, most existing approaches are naive and coarse-grained in means of either killing batch jobs to preempt resources for real-time jobs or reserving some resources in advance for real-time jobs to minimize the overtime ratio of real-time jobs, which prolongs the completion time of batch jobs and reduces the resource utilization of the cluster. In this paper, we propose UpPreempt, a fine-grained preemptive scheduling strategy for container-based clusters. When scheduling real-time jobs, UpPreempt considers the deadline of jobs and the resource usage of existing batch jobs. Performing resource preemption from multiple batch jobs is the basic idea of UpPreempt. The set of batch jobs to be preempted and the amount of resources taken from each job are finely determined. In this way, UpPreempt greatly alleviates the effect of preemption on the batch jobs in terms of completion time and avoids reserving resources for real-time jobs. We implement UpPreempt in YARN to evaluate its performance. The evaluation with various workloads shows that our proposed scheduling strategy can achieve a good trade-off between the overtime ratio of real-time jobs and the completion time of batch jobs without reducing the resource utilization of the cluster.

Real-time driving behavior monitoring is a corner stone to improve driving safety. Most of the existing studies on driving behavior monitoring using smartphones only provide detection results after an abnormal driving behavior is finished, not sufficient for driver alerting and avoiding car accidents. In this paper, we leverage built-in audio devices on smartphones to realize early recognition of inattentive driving events including Fetching Forward, Picking up Drops, Turning Back, and Eating or Drinking. Through empirical studies of driving traces collected in real driving environments, we find that each type of inattentive driving event exhibits unique patterns on Doppler profiles of audio signals. This enables us to develop an Early Recognition system, ER, which can recognize inattentive driving events at an early stage and alert drivers timely. ER employs machine learning methods to first generate binary classifiers for every pair of inattentive driving events, and then develops a modified vote mechanism to form a multi-classifier for all four types of inattentive driving events, for atypical inattentive driving events along with other driving behaviors. It next turns the multi-classifier into a gradient model forestto achieve early recognition of inattentive driving. Through extensive experiments with eight volunteers driving for about two months, ER can achieve an average total accuracy of 94.80 percent for inattentive driving recognition and recognize over 80 percent inattentive driving events before the event is 50 percent finished.

Auctions have been adopted by many major cloud providers, such as Amazon EC2. Unfortunately, only simple auctions have been implemented. Such simple auction has serious limitations, such as being unable to accept elastic user demands and having to allocate different types of VMs independently. These limitations create a big gap between the real needs of cloud users and the available services of cloud providers. In response to the limitations of the existing auction mechanisms, this paper proposes a novel online auction mechanism for IaaS clouds, with the unique features of an elastic model for inputting time-varying user demands and a unified model for requesting heterogeneous VMs together. However, several major challenges should be addressed, such as NP hardness of optimal VM allocation, time-varying user demands and potential misreports of private information of cloud users. We propose a truthful online auction mechanism for maximizing the profit of the cloud provider in IaaS clouds, which is composed of a price-based allocation rule and a payment rule. In the allocation rule, the online auction mechanism determines the number of VMs of each type to each user. In the payment rule, by introducing a marginal price function for each type of VMs, the mechanism determines how much the cloud provider should charge each cloud user. With solid theoretical analysis and trace-driven simulations, we demonstrate that our mechanism is truthful, fair and individually rational, and has a polynomial-time complexity. In addition, our auction achieves a competitive ratio for the profit of the cloud provider, compared against the offline optimal one.

Nowadays, an in-depth understanding of the Internet traffic profile is indispensable. Many network monitoring applications claim the connection-oriented analysis. However, many existing capture tools are too low-performance to cope with the large network traffic flow. Though some tools using traffic sampling can handle heavy network stream, there is information loss. While some parallel methods have been adopted to optimize the performance, there are insufficient extendible easy-use libraries. To feed this demand, this paper proposes a flexible parallel network library (Noff) which has the ability to process the network traffic at speed up to 6 Gbps. Noff supports two kinds of parallel processing procedures on the multi-core architecture to improve system compatibility and performance. Furthermore, Noff proposes a timing wheel scheme to process some network abnormity. For extendibility, Noff also offers many flexible callback interfaces. Some well-designed experimental evaluations indicate that Noff can perform about 25% better than other software libraries in terms of streams processing and can perform much better using a multi-core platform. Under the same circumstance, Noff can also process much higher concurrent traffic. Noff is currently deployed in a campus access point to process traffic at 500-800Mbps speed to offer stream information for a network security monitoring platform.

Benefiting from the pay-per-use pricing model of cloud computing, many companies migrate their services and applications from typical expensive infrastructures to the cloud. However, due to fluctuations in the workload of services and applications, making a cost-efficient VM configuration decision in the cloud remains a critical challenge. Even experienced administrators cannot accurately predict the workload in the future. Since the pricing model of cloud provider is convex other than linear that often assumed in past research, instead of typical scaling out strategy. In this paper, we adopt mix scale strategy. Based on this observation, we model an optimization problem aiming to minimize the VM configuration cost under the constraint of migration delay. Taking advantages of Lyapunov optimization techniques, we propose a mix scale online algorithm which achieves more cost-efficiency than that of scale out strategy. Experimental results shows that the mix scale algorithm saves 30.8% and 31.1% cost where controlling migration delay in a tolerable range under different workload respectively.

In the cloud computing architecture, virtual machine (VM) live migration is a fundamental research topic which has drawn extensive attention from communities of industry and academy. It is critical to transfer the VM memory pages that contain essential state information to resume the VM on another host during virtual machine (VM) live migration. There are many memory copy methods, such as pre-copy and post-copy. However, these methods have two limitations: application generality and performance imbalance. In this paper, we propose Ada-copy (adaptive copy), an adaptive memory copy strategy for VM live migration. The basic idea of Ada-copy is that the memory copy method of a VM should be determined by its workload characteristics. Specifically, based on the variation of current dirty page rate of memory, Ada-copy can adaptively select the most appropriate migration method to copy memory pages, thus addressing the two limitations of existing memory copy methods. To evaluate the effectiveness of our proposed strategy, we experiment with the Ada-copy on a variety of migration tasks with different dirty page rate and diverse memory usage workloads. Evaluation results show, compared with traditional methods, Ada-copy can significantly reduce the total migration time by 26%, the VM downtime by 42% and the amount of pages transferred by 35% in average.

Auctions have been adopted by many major cloud providers, such as Amazon EC2. Unfortunately, only simple auctions have been implemented. Such simple auction has serious limitations, such as being unable to accept elastic user demands and having to allocate different types of VMs independently. These limitations create a big gap between the real needs of cloud users and the available services of cloud providers. In response to the limitations of the existing auction mechanisms, this paper proposes a novel online auction mechanism for IaaS clouds, with the unique features of an elastic model for inputting time-varying user demands and a unified model for requesting heterogeneous VMs together. However, several major challenges should be addressed, such as NP hardness of optimal VM allocation, time-varying user demands and potential misreports of private information of cloud users. We propose a truthful online auction mechanism for maximizing the profit of the cloud provider in IaaS clouds, which is composed of a price-based allocation rule and a payment rule. In the allocation rule, the online auction mechanism determines the number of VMs of each type to each user. In the payment rule, by introducing a marginal price function for each type of VMs, the mechanism determines how much the cloud provider should charge each cloud user. With solid theoretical analysis and trace-driven simulations, we demonstrate that our mechanism is truthful and individually rational, and has a polynomial-time complexity.

Summary Shop recommendation in large shopping malls is useful in the mobile internet era. With the maturity of indoor positioning technology, customers' indoor trajectories can be captured by radio frequency identification devices readers, which provides a new way to analyze customers' potential preferences. In this paper, we design three methods for the top‐N shop recommendation problem. The first method is an improved matrix factorization method fusing estimated prior customer preference matrix that is constructed by Session‐based Temporal Graph computing. The second method is a Bayesian personalized ranking method based on the first method. The third method is by tensor decomposition combined with Session‐based Temporal Graph. Besides, we exploit customer history radio frequency identification devices trajectory information to find customers' frequent paths and revise predicted rating values to improve recommendation accuracy. Our methods are effective in modeling customers' temporal dynamics. At the same time, our approach considers repeated recommendation of the same shop by designing rating update rules. The test dataset is formed byJoyCitycustomer behavior records.JoyCityis a large‐scale modern shopping center in downtown Shanghai, China. The results show that our approaches are effective and outperform previous state‐of‐the‐art approaches. Copyright © 2016 John Wiley & Sons, Ltd.

To improve the performance of recommender systems in a practical manner, several hybrid approaches have been developed by considering item ratings and content information simultaneously. However, most of these hybrid approaches make recommendations based on aggregating different recommendation techniques using various strategies, rather than considering joint modeling of the item’s ratings and content, and thus fail to detect many latent factors that could potentially improve the performance of the recommender systems. For this reason, these approaches continue to suffer from data sparsity and do not work well for recommending items to individual users. A few studies try to describe a user’s preference by detecting items’ latent features from content-description texts as compensation for the sparse ratings. Unfortunately, most of these methods are still generally unable to accomplish recommendation tasks well for two reasons: (1) they learn latent factors from text descriptions or user--item ratings independently, rather than combining them together; and (2) influences of latent factors hidden in texts and ratings are not fully explored. In this study, we propose a probabilistic approach that we denote as latent random walk (LRW) based on the combination of an integrated latent topic model and random walk (RW) with the restart method, which can be used to rank items according to expected user preferences by detecting both their explicit and implicit correlative information, in order to recommend top-ranked items to potentially interested users. As presented in this article, the goal of this work is to comprehensively discover latent factors hidden in items’ ratings and content in order to alleviate the data sparsity problem and to improve the performance of recommender systems. The proposed topic model provides a generative probabilistic framework that discovers users’ implicit preferences and items’ latent features simultaneously by exploiting both ratings and item content information. On the basis of this probabilistic framework, RW can predict a user’s preference for unrated items by discovering global latent relations. In order to show the efficiency of the proposed approach, we test LRW and other state-of-the-art methods on three real-world datasets, namely, CAMRa2011, Yahoo!, and APP. The experiments indicate that our approach outperforms all comparative methods and, in addition, that it is less sensitive to the data sparsity problem, thus demonstrating the robustness of LRW for recommendation tasks.

To support mobile users in contributing sensing data for making urban management decisions, in ShanghaiGrid, unified data gathering operations are to be performed. For citywide coverage, public vehicles accept data from surrounding users and hand over to computing center through wireless base stations (BSs) deployed in the city. Meanwhile, several among the vehicles are hired as relays, which assist gathering from others with multicopy and multihop forwarding towards the BSs. However, the budget shared by deploying BSs and hiring relays is limited. We explore how to decide BS deployment and relay-based forwarding for efficient gathering under the budget. The challenge lies in the great uncertainty about collection opportunities of candidate locations and vehicles in future gathering processes. In this paper, we present an empirical approach for the problem. To tackle the challenges, we characterize collection performance as function of temporal data paths towards each candidate, and formulate the problem as a multiobjective optimization problem. To solve it, we reveal regular relations between the candidates and estimate expected importance of them with large set of real vehicular traces; and develop an algorithmic framework for BS deployment and corresponding forwarding strategy. Extensive trace-driven simulations demonstrate the efficacy of the approach.

This paper discusses the time-constrained data delivery problem in vehicular ad hoc networks (VANETs). The unique characteristics of the network present great challenges to the issue. First, there are no always-connected forwarding routes between vehicles. Second, there is an intrinsic tradeoff between communication cost and delivery quality. Third, there is great uncertainty about vehicular mobilities. Exploiting vehicular trajectories, we present a constructive approach called TNC to tackle the challenges. In TNC, contact-based data forwarding and communication connections through the mobile network are incorporated. TNC first predicts the time-stamped inter-vehicle contacts and establishes expected contact graph, based on which it then computes the configuration for enhancing the connectivity of the network while introducing minimum number of mobile communication connections. Extensive simulations based on three real vehicular traces collected from 2,000 taxis and 1,400 buses in Shanghai, and 2,500 taxis in Shenzhen have been conducted and results demonstrate the efficacy of our approach.

Recommending routes for a group of competing taxi drivers is almost untouched in most route recommender systems. For this kind of problem, recommendation fairness and driving efficiency are two fundamental aspects. In the paper, we propose SCRAM, a sharing considered route assignment mechanism for fair taxi route recommendations. SCRAM aims to provide recommendation fairness for a group of competing taxi drivers, without sacrificing driving efficiency. By designing a concise route assignment mechanism, SCRAM achieves better recommendation fairness for competing taxis. By considering the sharing of road sections to avoid unnecessary competition, SCRAM is more efficient in terms of driving cost per customer (DCC). We test SCRAM based on a large number of historical taxi trajectories and validate the recommendation fairness and driving efficiency of SCRAM with extensive evaluations. Experimental results show that SCRAM achieves better recommendation fairness and higher driving efficiency than three compared approaches.

Content-based publish/subscribe systems have been employed to deal with complex distributed information flows in many applications. It is well recognized that event matching is a fundamental component of such large-scale systems. Event matching searches a space which is composed of all subscriptions. As the scale and complexity of a system grows, the efficiency of event matching becomes more critical to system performance. However, most existing methods suffer significant performance degradation when the system has large numbers of both subscriptions and their component constraints. In this paper, we present Hash Tree (H-Tree), a highly efficient index structure for event matching. H-Tree is a hash table in nature that is a combination of hash lists and hash chaining. A hash list is built up on an indexed attribute by realizing novel overlapping divisions of the attribute's value domain, providing more efficient space consumption. Multiple hash lists are then combined into a hash tree. The basic idea behind H-Tree is that matching efficiencies are improved when the search space is substantially reduced by pruning most of the subscriptions that are not matched. We have implemented H-Tree and conducted extensive experiments in different settings. Experimental results demonstrate that H-Tree has better performance than its counterparts by a large margin. In particular, the matching speed is faster by three orders of magnitude than its counterparts when the numbers of both subscriptions and their component constraints are huge.

Event matching is the process of checking high volumes of events against large numbers of subscriptions and is a fundamental issue for the overall performance of a large-scale distributed publish/subscribe system. Most existing algorithms are based on counting satisfied component constraints in each subscription. As the scale of a system grows, these algorithms inevitably suffer from performance degradation. We present REIN (REctangle INtersection), a fast event matching approach for large-scale content-based publish/subscribe systems. The idea behind REIN is to quickly filter out unlikely matched subscriptions. In REIN, the event matching problem is first transformed into the rectangle intersection problem. Then, an efficient index structure is designed to address the problem by using bit operations. Experimental results show that REIN has a better matching performance than its counterparts. In particular, the event matching speed is faster by an order of magnitude when the selectivity of subscriptions is high and the number of subscriptions is large.

Recommending good driving paths is valuable to taxi drivers for reducing unnecessary waste in fuel and increasing revenue. Driving only according to personal experience may lead to poor performance. With the availability of large-scale GPS traces collected from urban taxis, we have the curiosity about whether we can discover the hidden knowledge in the trace data for smart driving recommendation. This paper focuses on developing a smart recommender system based on mining large-scale GPS trace datasets from a large number of urban taxis. However, such the trace datasets are in nature complex, large-scale, and dynamic, which makes mining the datasets particularly challenging. We first extract vehicular mobility pattern from the large-scale GPS trace datasets. Then, the optimal driving process is modeled as a Markov Decision Process (MDP). Solving the MDP problem results in the optimal driving strategy that gives smart recommendation for taxi drivers. In essence, the most rewarding driving paths can be derived in the long run. We have conducted extensive trace driven simulations and conclusive results show that our recommendation algorithm can successfully find good driving paths and outperforms other alternative algorithms.