Handover (HO) is one of the pivotal technologies for mobility management in highly dynamic mega LEO Earth Orbit satellite constellations (MLSCs). Due to the lack of channel reservation under random access (RA) and the prohibitive overhead incurred by centralized HO scheduling, intense competition among massive connections (e.g., IoRT nodes) results in significant degradation of service continuity. To solve the above challenges, we propose a distributed fairness-guided handover strategy (DHO-F) to dynamically select the best HO target and sub-channel. Specifically, the DHO-F optimization problem is formulated based on max-min egalitarian fairness and further modeled as a multi-objective Markov decision process (MOMDP), where fairness is expressed by the social welfare function (SWF). To address MOMDP, the analytical form of the policy gradient to maximize fairness is derived. Subsequently, Multi-agent Proximal Policy Optimization (MAPPO) with distributed cooperation is exploited to achieve long-term maximization. The fairness-guided MAPPO (FG-MAPPO) features a hybrid network architecture that simultaneously takes into account maximizing individual link rates and fairness among UEs. It reconciles these two conflicting objectives through the collaboration between a throughput-oriented (TO) network and a fairness-oriented (FO) network. Additionally, a distributed training framework is implemented to improve the sample efficiency and data diversity for on-policy FG-MAPPO. FG-MAPPO is fully compatible with 3GPP’s conditional handover (CHO) framework, demonstrating that it can be implemented in real-world. Extensive evaluations demonstrate that DHO-F demonstrates superior performance even compared to centralized algorithms, achieving an average improvement of 3.48% in SWF metric. Moreover, DHO-F establishes new SOTA performance in balancing fairness and rate maximization across medium-to-high load scenarios compared to IDQN, ISAC, and MAPPO.

Managing diverse data formats and improving query processing efficiency at cloud service centers is crucial in the era of interconnected heterogeneous devices. Lakehouse effectively manages heterogeneous data but faces challenges in maintaining SQL query performance and data independence in large-scale analyses. Precomputation, which stores intermediate results in advance, reduces query latency by a space-for-time trade-off. This paper addresses the challenges of matching and rewriting results in lakehouse environments. By combining pre-computation with the low-cost and dynamically scalable storage characteristics of cloud storage, we introduce an optimized lakehouse architecture with an in-memory index and dynamic task scheduling to enhance OLAP performance while balancing data center resources. Experiments demonstrate improvements in query performance ranging from 6.1 % to 64.9% over native lakehouse architectures and reductions in resource occupation from the original 75.6% to 17.3% with preloading and 25.0% with dynamic scheduling.

In non-terrestrial networks, satellite constellations based on the Low-Earth-Orbit (LEO) have become crucial for ensuring seamless global connectivity. The flexible continuity guarantee is demanded for task-oriented user connection requests in satellite networks. In this paper, we propose a task-oriented satellite conditional handover scheme, D-CHO, based on multi-agent game theory. For the problem formalization, this paper focuses on delay overhead, satellite utilization deviation, and load performance to model the multi-objective optimization. According to game theory, a Nash equilibrium exists among the multi-task game strategies that require satellite links. Through exploration and exploitation, the optimal satellite handover sequence scheme can be identified. This paper explores the optimal solution based on the MAPPO algorithm, which enables multiple tasks to make independent decisions based on their partial observations without requiring global information. This approach is beneficial for the adaptive expansion in response to dynamic changes in different satellite networks. The simulation results show that D-CHO improves performance by 22%, 26%, and 11% compared to the SCDP, G-CHO, and MADDPG-CHO algorithms, respectively, and exhibits better scalability while maintaining satisfactory performance.

The expansion of unmanned intelligent vehicles in daily living spaces increases the demand for their safe operation in human-machine coexistence environments. Effective space management is crucial to ensuring the safety and reliability of intelligent vehicles by preventing conflicts and controlling fear toward humans to a low level. Traditional spatial models, such as trajectory-based operation (TBO) and free-flight operation (FFO), are more or less deficient in maintaining low space occupancy and reducing the probability of conflicts. Hence, we present a MAPPO algorithm based on real-time dynamic motion data and a fear index model that quantifies human apprehension towards vehicles, aiming to allocate exclusive operational spaces for each vehicle. An octree-based spatial partitioning method is used to further mitigate conflicts among these space. Simulation experiments indicate that the proposed solution ensures collision-free allocation and reduced spatial occupancy that is 6.24% to 40.84% of other schemes , which achieves equilibrium between TBO and FFO.

Containers have become a critical component of cloud computing, offering lightweight virtualization for flexible deployment, elastic computing, and dynamic migration. However, the widespread adoption of the Overlay File System (OverlayFS) to provide file system services for containers faces significant I/O performance challenges in high-density and high-concurrency environments. The LRU page cache strategy is poorly suited to the memory access patterns of containers, and OverlayFS unnecessarily persists copy-on-write data. Furthermore, concurrent requests from multiple containers for host file system resources cause interference, further degrading performance on public cloud platforms. This paper proposes the Lifecycle-Aware Overlay File System (LaOvl), which leverages OverlayFS's redirection operations to accelerate container I/O in three key stages. LaOvl accurately senses the container lifecycle, enabling pre-loading of the page cache before startup, fine-grained temporary copy-on-write during runtime, and timely recycling of cold memory upon container destruction. Our evaluation, which includes micro-benchmarks and real-world applications, demonstrates that LaOvl significantly optimizes container I/O and achieves$\mathbf{9 5. 4 \%}$improvement over the original OverlayFS, outperforming state-of-the-art solutions.

With the launch of numerous low Earth orbit (LEO) satellites and the construction of constellation networks, the traditional bent-pipe communication method between LEO satellites and the ground has become inefficient, failing to fully utilize the satellites’ capabilities. To address this, we designed a model where satellites collaboratively execute complex missions at the space edge, reducing communication demands with the ground and improving task execution efficiency. Given the constraints of limited satellite resources, energy, and dynamic inter-satellite topology, we developed task scheduling and computation offloading algorithms using genetic algorithms to enhance task completion rates and resource utilization. Experimental results show that the proposed algorithms achieve good load balancing (approximately 10% of the greedy algorithm) and shorten scheduling length while minimizing timeouts.

With the continuous emergence of large-scale low-Earth orbit satellite constellations and the service-oriented functionality of satellites, we have designed a dynamic service substitution approach to enhance space-based missions’ reliability and completion rate. Based on our formal description model for satellite services, this approach generates one-to-one direct substitution or one-to-many composite substitution schemes when satellite service failures are detected, enabling task migration between satellites through real-time monitoring of satellite status. Additionally, when it is impossible to generate service substitution schemes that strictly meet function and resource requirements, this approach produces suboptimal solutions with relaxed constraints. These solutions are evaluated using our proposed evaluation model. Testing the prototype system demonstrated that our dynamic service substitution approach improves satellite missions’ reliability and completion rate compared to existing methods.

Singing Voice Conversion (SVC) is a technique that enables any singer to perform any song. To achieve this, it is essential to obtain speaker-agnostic representations from the source audio, which poses a significant challenge. A common solution involves utilizing a semantic-based audio pretrained model as a feature extractor However, the degree to which the extracted features can meet the SVC requirements remains an open question. This includes their capability to accurately model melody and lyrics, the speaker-independency of their underlying acoustic information, and their robustness for in-the-wild acoustic environments. In this study, we investigate the knowledge within classical semantic-based pretrained models in much detail. We discover that the knowledge of different models is diverse and can be complementary for SVC. Based on the above, we design a Singing Voice Conversion framework based on Diverse Semantic-based Feature Fusion (DSFF-SVC). Experimental results demonstrate that DSFF-SVC can be generalized and improve various existing SVC models, particularly in challenging real-world conversion tasks. Our demo website is available at https://diversesemanticsvc.github.io/.

Amphion is an open-source toolkit for Audio, Music, and Speech Generation, targeting to ease the way for junior researchers and engineers into these fields. It presents a unified framework that includes diverse generation tasks and models, with the added bonus of being easily extendable for new incorporation. The toolkit is designed with beginner-friendly workflows and pre-trained models, allowing both beginners and seasoned researchers to kick-start their projects with relative ease. The initial release of Amphion v0.1 supports a range of tasks including Text to Speech (TTS), Text to Audio (TTA), and Singing Voice Conversion (SVC), supplemented by essential components like data preprocessing, state-of-the-art vocoders, and evaluation metrics. This paper presents a high-level overview of Amphion. Amphion is open-sourced at https://github.com/open-mmlab/Amphion.

Crack segmentation is a pivotal task in assessing structural integrity across diverse domains, ranging from civil infrastructure such as bridges and buildings to the fabrication of heavy vehicles, which is crucial for ensuring the longevity and safety of materials. Despite its critical importance, the domain remains relatively underexplored within the academic sphere, particularly in accommodating the crack segmentation on resource-constrained devices. This challenge arises due to the inherent demand for deeper and broader network structures to achieve optimal performance, resulting in heavier computational and storage overhead. Thus, deploying crack segmentation models on practical platforms poses a formidable challenge. This paper presents a novel lightweight hybrid framework comprising robust Attention UNet architecture to assimilate comprehensive contextual information alongside the efficient MobileVit block to extract and integrate global contextual information utilizing the intricate self-attention mechanism. The intensive experiment results illustrate that our proposed method outperforms the existing state-of-the-art methods on the public benchmark datasets despite employing a reduced parameter space. The dataset and code can be accessed at: https://github.com/REINS-SJTU/TransAUnet

Graph active learning algorithms can reduce the amount of labeling and improve the applicability of graph neural networks. However, existing graph active learning algorithms are mainly performed at the node granularity. Those setting does not hold to datasets that are sensitive to edge attributes. To solve this problem, we propose a graph active learning algorithm at subgraph granularity. The algorithm tackles two critical challenges: how to estimate the expected labeling value of subgraphs and how to search for high-value subgraphs in the whole graph efficiently. For the first challenge, we evaluate the expected labeling values of subgraphs based on heuristic metrics, including uncertainty, representativeness, centrality, and diversity. Among them, uncertainty cannot be measured directly. Therefore, we measure subgraph inner cohesion by GNN attention weights and estimate uncertainty based on it. For subgraph search, we propose an efficient subgraph search algorithm. The proposed algorithm includes a simulated annealing search algorithm for a single subgraph and beam search with subgraph-effective reception field algorithms for multiple subgraphs. Experiments demonstrate that the subgraph granularity active learning algorithm proposed in this paper can achieve great results on edge-sensitive datasets.

With the rapid development of the MIoT(mobile Internet of things), the number of MIoT devices has increased rapidly. The collection and management of the status information continuously submitted by MIoT devices has brought great pressure to the network bandwidth of the data center. Existing edge data storage services lack the ability of data range query to support MIoT stream analysis, so we propose a cloud-edge collaborative data service, CEDS, to solve that problem. It uses edge computing nodes to store data sent by devices nearby and uses a central node to manage metadata and query data on edge nodes. In order to reduce the network transmission during data query, we use query splitting and pushdown optimization techniques to make each edge node only return aggregated data within the query range. To improve the query latency of global search, we propose an edge data indexing mechanism based on the compressed Rosetta filter. We test the CEDS performance with the taxi management task. Experiments illustrate that CEDS can efficiently support storing and range query of MIoT data streams with a small network traffic cost.

This paper investigates the long-term vessel trajectory prediction problem. A challenge in long-term prediction is modeling the navigation intention efficiently. We propose a novel model called Environment-Aware Vessel Trajectory Prediction Network (EA-VTP), which introduces the environment feature from the vessel density map. The vessel density map records the number of vessels of each position and therefore indicates the conventional tracks. A convolutional neural network is applied to the vessel density map to extract the information, which is then utilized by the recurrent module for the prediction. In addition, we introduce higher-order differentials and the residual prediction, which exploits the continuity of trajectory and balances the gradient of all steps. Besides, EA-VTP also provides the confidence range of position distribution, thus increases the reliability. The experiment shows that our EA-VTP outperforms baselines, and the environment feature effectively improves both short-term and long-term accuracy.

A vehicle-in-the-loop (ViL) testbench offers the possibility to test complex scenarios with ready-to-drive vehicles. For this purpose, the environmental sensors are simulated or stimulated. Essential component as a LiDAR is for automated driving systems (AD), its realistic behavior is hard to stimulate on the testbench. We propose a physics-based LiDAR model, which is real-time capable and shows many realistic features. This model simulates the important effects of laser propagation and reflection, mirror reflection motion distortion, reflection detectability and beam divergence. Besides that, we measured the reflectance of materials of interest to determine the reflection model parameters. Experiments proved that the simulation is real-time capable and the results showed a good match with measured data.

Hand actions classification is an imperative field for acquiring smart functionality in modern electronic devices because hand actions classification offers interactive and innovative methods to communicate and interact. Therefore, we develop a novel architecture based on you only looking at coefficients (YOLACT), a real-time instance segmentation approach, and a temporal relation network (TRN) for hand actions understanding. In addition, our framework consists of a face recognition-based security network (FRB-SN) for user identification. We trained the YOLACT and the TRN models using the segmented version of the 20BN jester dataset composed of hand actions images and ground truths while the FRB-SN is trained using the VGGFace2 dataset. For testing, the YOLACT is used to segment the object from the given image sequence and then passed to the TRN-trained model to predict the corresponding action. Our experimental results showed that the accuracy and frame rate of the proposed framework are competitive.

Human action recognition (HAR) is a popular topic in developing a visual analysis system because of its tremendous potential in autonomous visual analysis. However, visual analysis is a sophisticated field in computer vision because an image sequence consists of various features that do not belong to a specific action. Therefore, we present a novel architecture approach for human action recognition and localization. We dubbed it DTA, an abbreviation of the detect, track, and analyze. It is inspired by yolov3, deep-sort, and 3D convolutional neural networks. Our framework is compact in analyzing human action, and the results showed that the proposed method outperforms previous state-of-the-art methods in various aspects. Moreover, the action recognition model is developed, trained, and tested using the ROI version of the KTH dataset. The experimental results showed the accuracy of the proposed model is superior compared to other traditional methods.

Abstract Deep learning has assisted the field of single‐image super‐resolution (SR) in achieving new heights. However, the task of restoring a high‐resolution (HR) image from a highly degraded low‐resolution (LR) image is sophisticated due to poor image restoration quality. A novel and effective lightweight SR method is presented as super‐resolution via an enhanced feature block network (SREFBN) that successfully reconstructs an HR image using a corresponding LR image with a purposed deep residual block. In addition, a novel shared parameters approach in the top‐down pathway among low‐level feature maps is introduced. The experimental results prove that SREFBN achieves remarkable performance. The presented framework requires lower computational cost and outperforms many state‐of‐the‐art methods. It is also highly adaptable with low‐end devices, requiring lower multiplication and adding operations. A trade‐off comparison between the number of parameters, execution time, and accuracies is given while also showing different variations of our approach to prove the effectiveness and reliability of the shared parameters. Most importantly, the results indicate that our framework has gained state‐of‐the‐art performance on larger scales 3 and 4. Code is available at https://github.com/curzii23/SREFBN .

With the development of 5G technology and the proliferation of various mobile applications, mobile edge computing (MEC) provides services near the user side to meet the quality constraints of different tasks. Most current works focus on the offloading decision and resource allocation issues in MEC. However, few works focus on user mobility and the personalized preferences of different applications. In this paper, we study these issues and propose a deep reinforcement learning (DRL) based context-aware online offloading strategy. To further reduce the overhead caused by user mobility for task offloading and migration, we consider the user’s future movement trajectory and calculate the potential migration cost. Considering the dynamic network environment and the incompleteness of the observed system state information, we formulate the offloading decision problem as a partially observable Markov decision process (POMDP) problem, and then devise an efficient DRL algorithm to speed it up. We use EdgeCloudSim tool and Geolife trajectory to simulate the task offloading decision problem. The simulation results show that the proposed strategy is superior to other baseline strategies in terms of the total cost, delay, energy consumption, migration cost, and can be well adapted to different preferences and the dynamic network environment.

With the rapid growth of information, large amounts of graph-structured data have been generated. As an important task in graph-structured data research, node classification, which aims to classify nodes into different categories, has attracted a lot of attention from researchers in recent years. Real-life graphs are often dynamic whose graph topology and node attributes are constantly evolving. However, most of the studies focus on static graphs which can not capture the evolution of dynamic graphs. Node classification in dynamic graphs mainly has the following two challenges. First, it is difficult to effectively integrate modeling spatial and temporal features. Second, the evolution of dynamic graphs is located not only in node attributes but also in the graph topology. It is hard to learn the evolution of both aspects in the meantime. Besides, existing methods focus only on topological relations connected by explicit edges, while ignoring implicit topological relations that act in non-edge form. Implicit topological relations can help aggregate neighborhood features and further refine the modeling of node evolution patterns. To address these challenges and problems, we propose DS-TAGCN, a dual-stream topology attentive GCN for dynamic graph node classification. DS- TAGCN learns spatial-temporal features simultaneously by using a combination of GCN and LSTM. A dual-stream framework is designed to focus on the evolution of node attributes and graph topology, respectively. To mine the implicit topology, we propose TAGCN instead of GCN to model the implicit topological relations. Additionally, we incorporate a hierarchical attention mechanism in the network to automatically model the importance of different dimensional features. Extensive experiments demonstrate the effectiveness of DS-TAGCN.

With the unprecedented traffic demand from various mobile devices, bad quality of experience arises in traditional reactive networks, such as long loading time and frozen in the middle. This paper presents Predict-then-Prefetch caching strategy in 5G networks to improve the quality of experience. This strategy partitions the capacity of the base stations into the proactive cache to prefetch popular content for a sum total maximum of popularity and the reactive one to cache content which is unpopular or whose popularity can't be forecast inaccurately. It is demonstrated that Predict-then-Prefetch caching strategy has the best proportion of the proactive cache with different percentages of time-related content. Under this best proportion of the circumstances where all content is time-related, this strategy improves hit ratio by 30% and reduces latency by 50% in the architecture of 200M small base stations, which could enhance the quality of experience to a great degree.

Because of the popularity of cloud computing, Cloud Service Providers (CSPs) can rent virtual machines (VMs) from Cloud Providers (CPs) conveniently. In our previous work, we proposed an autonomic and elastic resource scheduling framework, named AERS, which made full use of both proactive and reactive controllers in the field of dynamic resource provision and was integrated with an availability-aware and communication overhead optimized placement strategy. In this paper, we propose an enhanced version of AERS, named EAERS. It eliminates modeling the specific cloud application and instead determines the relationship between workloads and the number of virtual machines (VMs) through self-learning so that the whole scheduling can be carried out in a more transparent manner. Dynamic consolidation is also designed, implemented and integrated into EAERS. Experiments on OpenStack show that EAERS performs as expected.

With no limit on time and location [1], the number of users attracted by massive open online course (MOOC) has increased rapidly, and many platforms have been built to provide a variety of courses. All of these trigger an explosive growth in data volume. As we known, people have met big data in many areas and proposed many techniques and methods to deal with them. However, people still have no sense of the data system in MOOC. How to dig out valuable information from its big data to benefit all stockholders is an instant challenge. This paper has an insight into the data system in MOOC and proposes Data Ming goals and methods. At last, the paper will introduce the computing platform and CNMOOC as a case study.

Battery systems are crucial components for mission-critical data centers. Without secure energy backup, existing under-provisioned data centers are largely unguarded targets for cyber criminals. Particularly for today's scale-out servers, power oversubscription unavoidably taxes a data center's backup energy resources, leaving very little room for dealing with emergency. Besides, the emerging trend towards deploying distributed energy storage architecture causes the associated energy backup of each rack to shrink, making servers vulnerable to power anomalies. As a result, an attacker can generate power peaks to easily crash or disrupt a power-constrained system. This study aims at securing data centers from malicious loads that seek to drain their precious energy storage and overload server racks without prior detection. We term such load as Power Virus (PV) and demonstrate its basic two-phase attacking model and characterize its behaviors on real systems. The PV can learn the victim rack's battery characteristics by disguising as benign loads. Once gaining enough information, the PV can be mutated to generate hidden power spikes that have a high chance to overload the system. To defend against PV, we propose power attack defense (PAD), a novel energy management patch built on lightweight software and hardware mechanisms. PAD not only increases the attacking cost considerably by hiding vulnerable racks from visible spikes, it also strengthens the last line of defense against hidden spikes. Using Google cluster traces we show that PAD can effectively raise the bar of a successful power attack: compared to prior arts, it increases the data center survival time by 1.6~11X and provides better performance guarantee. It enables modern data centers to safely exploit the benefits that power oversubscription may provide, with the slightest cost overhead.

Big data analysis has gained significant popularity within the last few years. The MapReduce framework presented by Google makes it easier to write applications that process vast amount of data. MapReduce targets at large commodity clusters where failures are not exceptions. However, Hadoop, the most popular implementation of MapReduce performs poorly under failures. Hadoop implements the fault tolerance strategy at the task level, as a result, a task failure will require a re-execution of the whole task regardless of how much input has already been processed. In this paper, we present BeTL which introduces slight changes to the execution flow of MapReduce, and makes it possible to gain a finer-grained fault tolerance. Map tasks can create checkpoints so that a retrying task doesn't have to start from scratch and thus saves much time. Speculation strategy can also benefit from this. The new execution flow involves less IO operations and performs better than Hadoop even under no failures. In our experiments, BeTL outperforms Hadoop by 6.6 percent on average under no failures and 4.6 to 51.0 percent under different failure densities.

Massive open online course is now a popular choice for online learners. There are many MOOC platforms all over the world. They provide multiple variants of MOOC. Traditionally, students learn by watching videos and doing online quizzes. Course forum is also an important component of MOOC. It is a good place for opinions sharing and discussion. This paper focuses on the relationship between students' academic performance and their participation in the course forum. It also studies the semantics of both students and TAs' posts in the course forum in order to understand their behavior in the forum. We found those who achieve higher scores tend to be more active in forum. But they also write a higher percentage of posts that are unrelated to the course than those who get lower scores. TAs are not very active in the forum and they have limited impacts. This paper talks about the problems of current course forum participation and presents some suggestions for MOOC forum.

MapReduce is a programming paradigm that makes it simple and efficient to process vast amount of data. It targets at very big clusters, where failures are no longer exceptions. Fault tolerance is vital to MapReduce, however, fault tolerance and recovery strategies in MapReduce perform poorly under failures. Currently fault tolerance is implemented at the task level, a task failure will lead to a re-execution of the whole task. In this work, we present ReCT, a family of resilient checkpointing tactics(ReCT) to intensively improve MapReduce performance under map task failures. ReCT introduces slight changes to current MapReduce execution flow and makes it possible to create checkpoints beneath the task level. In case of task failures, ReCT tries to make the most of finished partial tasks and skip them in retry attempts. The checkpointing tactics bring little overhead and intensively accelerate fault recovery process. We also observe that under some circumstances, the new execution flow in ReCT involves much less IO operations than that in Hadoop. ReCT outperforms Hadoop by 6.6% on average under no failures and 4.6% to 51.0% under different failure densities.

This paper presents the design, implementation, and evaluation of a multi-tiered storage system called MOBBS, which provides hybrid block storage for Virtual Machines (VMs) on top of object-based storage infrastructure. MOBBS is mainly motivated by the gap between the lack of studies on hybrid block storage for VMs and the increasing prevalence of hybrid storage systems. By stripping disk images into partitions and intelligently storing them on different storage tiers according to real-time workload patterns, MOBBS achieves efficient use of multiple storage devices and relieves the burden of data placement. Leveraging the benefits of object-based storage, MOBBS is able to dynamically perform non-disruptive and fine-grained data migration between storage tiers and distribute the complexity of data migration across entire storage nodes. Such designs enable our system to deliver storage for VMs with high scalability and availability under an efficient use of SSDs. We evaluated a Ceph implementation of MOBBS using both block and file system workloads. The results comprehensively demonstrate MOBBS's effectiveness in performance improvement as well as efficient utilization of different storage devices.

As a representative of NO-SQL database, MongoDB is widely preferred for its automatic load-balancing to some extent, which including distributing read load to secondary node to reduce the load of primary one and auto-sharding to reduce the load onspecific node through automatically split data and migrate some ofthem to other nodes. However, on one hand, this process is storage-load -- Cbased, which can't meet the demand due to the facts that some particular data are accessed much more frequently than others and the 'heat' is not constant as time going on, thus the load on a node keeps changing even if with unchanged data. On the other hand, data migration will bring out too much cost to affect performance of system. In this paper, we will focus on the mechanism of automatic load balancing of MongoDB and proposean heat-based dynamic load balancing mechanism with much less cost.

Cloud computing has attracted increasing attention in recent years. With the growth in the number and frequency of applications being deployed into clouds, the burden of resource management of cloud providers is becoming heavier. The resource deployment must satisfy the need about the performance, availability and reliability of applications from the view of clients, but also ensure the high resource utilization of the cloud providers. In this paper, we design a multi-objective serial optimization with priorities approach, named RMORM, to find the resource deployment in clouds rapidly. This approach is of great practical significance and engineering value and scalable to add new constraints.

With object storage systems being increasingly recognized as a preferred way to expose one's storage infrastructure to the web, the past few years have witnessed an explosion in the acceptance of these systems. Unfortunately, the proliferation of available solutions and the complexity of each individual one, coupled with a lack of dedicated workload, makes it very challenging for one to evaluate and tune the performance of different systems. To help address this problem, we present the Cloud Object Storage Benchmark (COSBench). It is a benchmark tool that we have developed at Intel with the goal of facilitating both performance comparison and system optimization of these systems. In this paper, we describe the design and implementation of this tool, focusing on its extensibility and scalability. In addition, we discuss how people can use this tool to perform system characterization and how the latter can facilitate system comparison and optimization. To demonstrate the value of our tool, we report the results of our experiments conducted on two Swift setups we built in our lab. We also share some of our experiences in turning our setups to achieve higher performance.

The availability of Web applications influenced by Virtual Machine (VM)-based physical locations during resource scaling is a crucial concern for customers and cloud providers. In this paper, we present a novel computing model to describe availability attribute of one application in hierarchical structured cloud. Meanwhile, we propose an availability-aware approach to explore how and where to allocate computing resource via vertical and horizontal scaling. Partial experimental results in simulation environment are also presented.

With object storage services becoming increasingly accepted as replacements for traditional file or block systems, it is important to effectively measure the performance of these services. Thus people can compare different solutions or tune their systems for better performance. However, little has been reported on this specific topic as yet. To address this problem, we present COSBench (Cloud Object Storage Benchmark), a benchmark tool that we are currently working on in Intel for cloud object storage services. In addition, in this paper, we also share the results of the experiments we have performed so far.

Cloud Federation is one of the ideal solutions tothis random burst traffic problem. It focuses on 'borrowing'computing resources from foreign Clouds when home Cloud isabout to overloaded and 'leasing' resources to foreign Cloudswhen home Cloud is free. Considering crucial importance ofbusiness value for a Cloud provider, we need to perform somedynamic, reasonable and simple rules or mechanisms to helpCloud provider making better billing decisions and improvingthe overall performance in a federation scope. Cloud Federationis implemented by network connections, so in common situationthe best candidate Cloud to be federated enjoys a higher speedof connection to the home Cloud. This remains a problem thatseveral Clouds located in a subnet have a high possibility to befederated with each other, thus complicated dependency relationsamong them will appear. In this paper, we have carried out aseries of mechanisms involving dynamic resource arrangement inestablishment and deconstruction of a Cloud Federation, in orderto sort out these dependency relations which are the potentialrisk factors to the overall performance.

Hadoop has become one popular framework to process massive data sets in a large scale cluster. However, it is observed that the detection of the failed worker is delayed, which may result in a significant increase in the completion time of jobs with different workload. To cope with it, we present two mechanisms: Adaptive interval and Reputation-based Detector that support Hadoop to detect the failed worker in the shortest time. The Adaptive interval is trying to dynamically configure the expiration time which is adaptive to the job size. The Reputation-based Detector is trying to evaluate the reputation of each worker. Once the reputation of a worker is lower than a threshold, then the worker will be considered as a failed worker. In our experiments, we demonstrate that both of these strategies have achieved great improvement in the detection of the failed worker. Specifically, the Adaptive interval has a relatively better performance with small jobs, while the Reputation-based Detector is more suitable for large jobs.

In order to improve the efficiency of service discovery and release the load of service registry, this paper proposes a service registry model named as SRMC (Service Registry with Multilevel Caches) which clusters the service consumers into groups according to their searching similarity and sets up a multilevel cache for all groups to improve the performance of service discovery. The multilevel caches of SRMC are refreshed by a hybrid mechanism which includes event-based refreshing and periodical refreshing. The basis of refreshing and clustering is the history records of service discovery requests issued by service consumers. The running results of an instance of SRMC deployed in an experimental environment have shown that SRMC is effective to reduce the times of accessing global storage and the amount of data searched in service discovery.

Service-Oriented Architecture (SOA) provides a flexible framework for service composition. In a service market scenario, given a functional description of service, different providers may offer diverse service implementations that match such a functional description, but differ for some QoS attributes. It is increasingly vital to provide a service selection and recommendation mechanisms that best meet the QoS requirements of the service user. Different from most of the existing approaches to service selection, we consider a Web service selection and ranking mechanism with multi-QoS attributes, focusing on simulating degree of consumer satisfaction and hypothesizing consumer preference historical information. Efficient service selection mechanism and heuristic algorithm for consumer preference of multi-QoS are presented in this article and their performances are studied by simulations.

SOFL (structured object-oriented formal language) is a formal engineering language and method for software system analysis, specification and design. It has been used in many systems. But its limitations result in the lack of the support for the modeling of middleware-based transaction management. However, the transaction management has been the necessary feature of distributed applications. So we extend the SOFL features to enable SOFL to support modeling of middleware-based transaction management by adding structures of CDFD and syntax of specification.

SOFL is a formal language and method for software system analysis, specification and design, and it fully supports structured techniques and object-oriented techniques. AOP (aspect-oriented programming) is a new technique for software development. Since AOP leads a completely different way from structured or object-oriented techniques, original SOFL can not been used for AOP modeling. In this paper, we extend SOFL and introduce several new features which will enable SOFL to be used to fully and clearly specify AOP features, such as aspects, pointcuts and advices.