In recent years, volumetric videos have gradually prospered as an intriguing video paradigm, offering users a fully immersive viewing experience with six Degrees of Freedom (DoF). However, most current live volumetric video streaming methods struggle to facilitate the real-time performance requirements due to the nature of frequent user interactions and the complexity of network environments during video playback. Inspired by the correlation between the human visual effects and adjacent frame motion features, we proposeCetus, a context-aware cross-layer coordination system for live volumetric videos. First, we present an application-layer Neural Radiance Fields (NeRF)-based codec framework that leverages spatio-temporal semantic information for optimizing the compression quality of each video frame. Second, we exploit a flexible cross-layer coordination framework that seamlessly integrates frame drop strategy with partially reliable transmission, orchestrating transport protocols and application-informed rates to enhance the Quality of Experience (QoE) for multiple users. Furthermore, we develop a lightweight branching decision tree algorithm that adaptively makes fine-grained frame drop decisions. Experimental evaluations of our implemented system prototype demonstrate that Cetus significantly outperforms existing baseline approaches. Compared to the state-of-the-art baselines, Cetus effectively improves video frame rate by at least 24.7% and video quality by an average of 32.6%.

The convergence of IoT, edge, cloud, and HPC technologies creates a heterogeneous compute continuum requiring sophisticated workload management. Current tools like SLURM, Kubernetes, and Snakemake lack automated optimization for cross-platform resource allocation, forcing users to manually map workloads across diverse infrastructures. We present a comprehensive framework integrating heterogeneous system and workload modeling integration with Snakemake followed by different tools and techniques like Mixed Integer Linear Programming (MILP) for multi-objective optimization to automate task mapping and scheduling across the compute continuum. Our approach extends Snakemake scheduler with formal mathematical models that optimize resource utilization and makespan. Experimental evaluation demonstrates that MILP-based solution achieves optimal scheduling for small-scale workflows (5x5 tasks) in 0.02 seconds, while heuristic methods provide 99.9% faster solutions for large-scale scenarios (5000×5000 tasks) with only 5-10% deviation from optimal makespan. For parallel workflows, the optimization achieves up to 16.7% makespan reduction compared to sequential scheduling approaches.

Vision Language Models (VLMs) have undergone significant advancements, particularly with the emergence of mobile-oriented VLMs, which offer a wide range of application scenarios. However, the substantial computational requirements for training these models present a significant obstacle to their practical application. To address this issue, Low-Rank Adaptation (LoRA) has been proposed. Nevertheless, the standard LoRA with a fixed rank lacks sufficient capability for training mobile VLMs that process both text and image modalities. In this work, we introduce HyDRA, a parameter-efficient fine-tuning framework designed to implement hierarchical and dynamic rank scheduling for mobile VLMs. This framework incorporates two essential optimization strategies: (1) hierarchical optimization, which involves a coarse-grained approach that assigns different ranks to various layers, as well as a fine-grained method that adjusts ranks within individual layers, and (2) dynamic adjustment, which employs an end-to-end automatic optimization using a lightweight performance model to determine and adjust ranks during the fine-tuning process. Comprehensive experiments conducted on popular benchmarks demonstrate that HyDRA consistently outperforms the baseline, achieving a 4.7% improvement across various model sizes without increasing the number of trainable parameters. In some tasks, it even surpasses full-parameter fine-tuning.

Federated Identity Management (FIM) allows different domains or organizations to use the same identification data. With FIM, an Identity Provider (IdP) can provide attributes or credentials to a Service Provider (SP) based on a trust relationship. Currently, these attributes are primarily stored and managed by IdPs, with users having limited control. In contrast, Self-Sovereign Identity (SSI) empowers users to manage their verified credentials independently of a central authority. To address this limitation while considering SSI’s functionality, this paper proposes a prototype that enables FIM to request verified attributes or credentials from SSI users by integrating part of the FIM flow into the SSI verification process without replacing the existing authentication process on the IdP. To evaluate this prototype, we developed a module within the Keycloak application server that integrates attribute requests in the FIM flow with the SSI verification process. In addition, we conducted performance tests of attribute verification in this flow using two DID methods (did:sov and did:web) with 25–200 virtual users.

In this paper, we introduced a novel methodology to build a classification‐forecast model used for financial risk forewarning. For the first step, we utilize the K–S test, Mann–Whitney U test, and Pearson’s correlation to select variables. Then, we employ CRITIC and fuzzy comprehensive evaluation (FCE) methods to score the risk of listed companies. Following this, self‐organizing maps (SOM) clustering is utilized to segment the samples into five distinct risk levels. For the second step, we utilized triangulated maximally filtered graph (TMFG) and maximally filtered clique forest (MFCF) to minimize the number of indicators based on the dependent relationships between variables. These are then combined with Gaussian Markov random field (GMRF) and Copula algorithms to address nonlinear situations, forming what we refer to as the LoGo model. To further enhance the accuracy of LoGo models, we utilize the square Mahalanobis distance to compute the log‐likelihoods as part matrix. The results reveal that the enhanced LoGo model with part matrix improves average accuracy by 7% compared with the original models without part matrix, albeit with a tenfold increase in execution time. MFCF demonstrates superior performance over TMFG in linear situations, achieving a 40% higher accuracy. However, under nonlinear circumstances, TMFG only requires half the execution time of MFCF, yet achieves a slightly higher average accuracy. Furthermore, compared with the widely used CNN models, the enhanced LoGo models show superior performance as they achieved closed accuracy in a shorter time.

FEVEROUS is a benchmark and research initiative focused on fact extraction and verification tasks involving unstructured text and structured tabular data. In FEVEROUS, existing works often rely on extensive preprocessing and utilize rule-based transformations of data, leading to potential context loss or misleading encodings. This paper introduces a simple yet powerful model that nullifies the need for modality conversion, thereby preserving the original evidence’s context. By leveraging pre-trained models on diverse text and tabular datasets and by incorporating a lightweight attention-based mechanism, our approach efficiently exploits latent connections between different data types, thereby yielding comprehensive and reliable verdict predictions. The model’s modular structure adeptly manages multi-modal information, ensuring the integrity and authenticity of the original evidence are uncompromised. Comparative analyses reveal that our approach exhibits competitive performance, aligning itself closely with top-tier models on the FEVEROUS benchmark.

Herein, we presented a novel methodology for constructing nonlinear parsimonious probabilistic models (Copula–LoGo). Our approach combines Gaussian–Copula with linear LoGo probabilistic models to obtain a nonlinear coefficient matrix, which cannot be accurately estimated using Pearson’s correlation coefficients or kernel tricks. Using the Gaussian–Copula obtained from a multivariate normal distribution through the probability integral transform, we applied the triangulated maximally filtered graph algorithm to construct decomposable information filtering networks. By analyzing the topological structure of the networks, we discerned the dependence characteristics for each variable in the regression model. Subsequently, by integrating the Gaussian Markov random field, we computed the estimate of the global sparse inverse covariance J from subparts of the dependence network and constructed the nonlinear, parsimonious model. Thus, our methodology exhibits computational efficiency and statistical robustness, outperforming other regression methods such as Support Vector Machine, Long Short-Term Memory, Principal Components Analysis, and XGBoost models, particularly in the financial domain, as evidenced by improved performance metrics including mean absolute error, root mean square error, and R2 score (prediction accuracy) and reduced execution time.

Abstract This study addresses the issue of power consumption in virtualized cloud data centers by proposing a virtual machine (VM) replacement model and a corresponding algorithm. The model incorporates multi-objective functions, aiming to optimize VM selection based on weights and minimize resource utilization disparities across hosts. Constraints are incorporated to ensure that CPU utilization remains close to the average CPU usage while mitigating overutilization in memory and network bandwidth usage. The proposed algorithm offers a fast and efficient solution with minimal VM replacements. The experimental simulation results demonstrate significant reductions in power consumption compared with a benchmark model. The proposed model and algorithm have been implemented and operated within a real-world cloud infrastructure, emphasizing their practicality.

Linking digital accounts belonging to the same user has progressed from a research topic to a foundation for security, user satisfaction, and developing next-generation services. Still, few studies address account linkage in domains other than social networks. This deficiency is particularly apparent in federated domains such as academia, where network-based information and contextual data are typically unavailable. To address this issue, we propose SmartSSO, a framework that aims to automate the account linkage process by analyzing user routines and behavior during login processes. SmartSSO adapts two self-attention-based models to generate new representations from user patterns in a lower-dimensional latent space where the learned structure is employed to identify related accounts held by a user. We show that the trained models on a large corpus of production data, including more than one million samples gathered over six months from 50,000 users, achieve over 98% accuracy in hit-precision.

Abstract This study addresses the critical need for effective resource management through software container migration in cloud data centers. It emphasizes its role in avoiding resource shortages and reducing energy consumption in cloud environments. This study introduces a novel multi-objective integer linear programming (ILP) approach for software container replacements, complemented by a specialized algorithm designed to migrate software containers between over- and underutilized hosts to enhance efficiency compared to traditional optimization methods. The simulation results demonstrate the algorithm's effectiveness, validating its potential for optimizing energy usage and resource allocation in cloud environments. Statistical analyses confirm the proposed model's and algorithm's superiority over benchmark approaches, highlighting their potential for enhancing resource management in cloud computing systems.

In the traditional video streaming service provisioning paradigm, users typically request video contents through nearby Content Delivery Network (CDN) server(s). However, because of the uncertain wide area networks delays, the (remote) users usually suffer from long video streaming delay, which affects the quality of experience. Multi-Access Edge Computing (MEC) offers caching infrastructures in closer proximity to end users than conventional Content Delivery Networks (CDNs). Yet, for video caching, MEC's potential has not been fully unleashed as it overlooks the opportunities of collaborative caching and multi-bitrate video transcoding. In this paper, we model and formulate an Integer Linear Program (ILP) to capture the long-term cost minimization problem for caching videos at MEC, allowing joint exploitation of MEC with CDN and real-time video transcoding to satisfy arbitrary user demands. While this problem is intractable and couples the caching decisions for adjacent time slots, we design a polynomial-time online orchestration framework which first relaxes and carefully decomposes the problem into a series of subproblems solvable in each individual time slot and then converts the fractional solutions into integers without violating constraints. We have formally proved a parameterized-constant competitive ratio as the performance guarantee for our approach, and also conducted extensive evaluations to confirm its superior practical performance. Simulation results demonstrate that our proposed algorithm outperforms the state-of-the-art algorithms, with 13.6% improvement on average in terms of total cost.

Network Function Virtualization (NFV) has been emerging as an appealing solution that transforms from dedicated hardware implementations to software instances running in a virtualized environment. In NFV, the requested service is implemented by a sequence of Virtual Network Functions (VNF) that can run on generic servers by leveraging the virtualization technology. These VNFs are pitched with a predefined order, and it is also known as the Service Function Chaining (SFC). Considering that the delay and resiliency are two important Service Level Agreements (SLA) in a NFV service, in this paper, we first investigate how to quantitatively model the traversing delay of a flow in both totally ordered and partially ordered SFCs. Subsequently, we study how to calculate the VNF placement availability mathematically for both unprotected and protected SFCs. After that, we study the delay-sensitive Virtual Network Function placement and routing problem with and without resiliency concerns. We prove that this problem is NP-hard under two cases. We subsequently propose an exact Integer Nonlinear Programming (INLP) formulation and an efficient heuristic for this problem in each case. Finally, we evaluate the proposed algorithms in terms of acceptance ratio, average number of used nodes and total running time via extensive simulations.

Network Function Virtualization (NFV) has been emerging as an appealing solution that transforms complex network functions from dedicated hardware implementations to software instances running in a virtualized environment. Due to the numerous advantages such as flexibility, efficiency, scalability, short deployment cycles, and service upgrade, NFV has been widely recognized as the next-generation network service provisioning paradigm. In NFV, the requested service is implemented by a sequence of Virtual Network Functions (VNF) that can run on generic servers by leveraging the virtualization technology. These VNFs are pitched with a predefined order through which data flows traverse, and it is also known as the Service Function Chaining (SFC). In this article, we provide an overview of recent advances of resource allocation in NFV. We generalize and analyze four representative resource allocation problems, namely, (1) the VNF Placement and Traffic Routing problem, (2) VNF Placement problem, (3) Traffic Routing problem in NFV, and (4) the VNF Redeployment and Consolidation problem. After that, we study the delay calculation models and VNF protection (availability) models in NFV resource allocation, which are two important Quality of Service (QoS) parameters. Subsequently, we classify and summarize the representative work for solving the generalized problems by considering various QoS parameters (e.g., cost, delay, reliability, and energy) and different scenarios (e.g., edge cloud, online provisioning, and distributed provisioning). Finally, we conclude our article with a short discussion on the state-of-the-art and emerging topics in the related fields, and highlight areas where we expect high potential for future research.

Summary With the popularity of mobile devices (such as smartphones and tablets) and the development of the Internet of Things, mobile edge computing is envisioned as a promising approach to improving the computation capabilities and energy efficiencies of mobile devices. It deploys cloud data centers at the edge of the network to lower service latency. To satisfy the high latency requirement of mobile applications, virtual machines (VMs) have to be correspondingly migrated between edge cloud data centers because of user mobility. In this paper, we try to minimize the network overhead resulting from constantly migrating a VM to cater for the movement of its user. First, we elaborate on two simple migration algorithms (M‐All and M‐Edge), and then, two optimized algorithms are designed by classifying user mobilities into two categories (certain and uncertain moving trajectories). Specifically, a weight‐based algorithm (M‐Weight) and a mobility prediction–based heuristic algorithm (M‐Predict) are proposed for the two types of user mobilities, respectively. Numerical results demonstrate that the two optimized algorithms can significantly lower the network overhead of user mobility–induced VM migration in mobile edge computing environments.

One of the most challenging obstacles for the advancement of clouds is the lack of assurance and transparency, along with the current paucity of techniques to quantify security. A fundamental requirement for solving this problem is to provide proper levels of security based on the requirements of cloud customers and sensitivity of data. However, generating these security levels only partially serves the requirements of cloud customers especially if it's not linked to the management of SLC commitments. Accordingly, a novel approach has been proposed in this paper to define and manage security indicators in the generated security levels. These indicators are used to react to eventualities that may threaten the established mechanisms of the generated security rings, to ensure the fulfillment of agreed assurance levels, and to minimize the damages in case of attacks, unpredictable events or unavoidable changes. The proposed schema uses simultaneous monitoring with two different stand-alone agents per security level. These agents are initialized based on all security policies in the SLA to enhance the process of monitoring and rectification and to increase the rate of satisfaction and reliability in clouds.

For many application scenarios, interconnected data centers provide high service flexibility, reduce response time, and facilitate timely data backup. Many data center system parameters might have variant impact on the interconnection performance. Despite many studies on data center network performance, there exist few analytical work that reveal insightful knowledge with wide range of system parameters as input, especially focusing on data center interconnects (DCI). This paper creates analytical models for representative data center network architectures and provides the performance calculus aiming to apply for data center interconnects. By parameterising the number of devices, the arriving traffics, the switch link capacities, and the traffic locality, we derive the relationship among the DCI bandwidth, inter-DC latency, and these parameters. Based on this, further discussion and numerical examples investigate and evaluate the modelling and calculus from multiple angles and show the possibility how this calculus assists DC/DCI design and operation.

The rapid popularity of mobile devices especially smart phones has changed human life style greatly. In this paper, we examine the consumer behaviors on several e-commerce platforms based on a large-scale dataset of mobile internet access records for about 3.5 months from a major telecom operator in China, which covers 126,388 users from Shanghai. We provide a preliminary study on users' daily and periodic online shopping behaviors, as well as the influence of special online shopping events and gender factors. These findings may be exploited by e-commerce providers e.g., for developing personalized recommendation systems to improve their service quality and profit.

Live Virtual Machine (VM) migration offers a couple of benefits to cloud providers and users, but it is limited within a data center. With the development of cloud computing and the cooperation between data centers, live VM migration is also desired across data centers. Based on a detailed analysis of VM deployment models and the nature of VM image data, we design and implement a new migration framework called CBase. The key concept of CBase is a newly introduced central base image repository for reliable and efficient data sharing between VMs and data centers. With this central base image repository, live VM migration and further performance optimizations are made possible. The results from an extensive experiment show that CBase is able to support VM migration efficiently, outperforming existing solutions in terms of total migration time and network traffic.

The concept of persistent identification is increasingly important for research data management.At the beginnings it was only considered as a persistent naming mechanism for research datasets, which is achieved by providing an abstraction for addresses of research datasets.However, recent developments in research data management have led persistent identification to move towards a concept which realizes a virtual global research data network.The base for this is the ability of persistent identifiers of holding semantic information about the identified dataset itself.Hence, community-specific representations of research datasets are mapped into globally common data structures provided by persistent identifiers.This ultimately enables a standardized data exchange between diverse scientific fields.Therefore, for the immense amount of research datasets, a robust and performant global resolution system is essential.However, for persistent identifiers the number of resolution systems is in comparison to the count of DNS resolvers extremely small.For the Handle System for instance, which is the most established persistent identifier system, there are currently only five globally distributed resolvers available.The fundamental idea of this work is therefore to enable persistent identifier resolution over DNS traffic.On the one side, this leads to a faster resolution of persistent identifiers.On the other side, this approach transforms the DNS system to a data dissemination system.

Customers often suffer from the variability of data access time in cloud storage service, caused by network congestion, load dynamics, etc. One solution to guarantee a reliable latency-sensitive service is to issue requests with multiple download/upload sessions, accessing the required data (replicas) stored in one or more servers. In order to minimize storage costs, how to optimally allocate data in a minimum number of servers without violating latency guarantees remains to be a crucial issue for the cloud provider to tackle. In this paper, we study the latency-sensitive data allocation problem for cloud storage. We model the data access time as a given distribution whose Cumulative Density Function (CDF) is known, and prove that this problem is NP-hard. To solve it, we propose both exact Integer Nonlinear Program (INLP) and Tabu Search-based heuristic. The proposed algorithms are evaluated in terms of the number of used servers, storage utilization and throughput utilization.

The immense growth of digital data has launched a series of profound changes in today's world. In order to handle the huge volume of data, the underlying systems are steadily subjected to optimizations. Often these systems in turn are composed of as complex multi-tier systems, whereas each is tasked with a specific function on the incoming dataset. However, the increasing complexity of these multi-tier internet systems requires appropriate techniques to estimate the benefit and the impact of costly improvement endeavors. This paper provides a mathematical basis to investigate the effects of individual tier improvements to the overall speedup of multi-tier internet based systems. The fundamental approach in this paper is to analyze the behavior of the well-known Mean-Value-Algorithm (MVA), which is an established prediction tool for computer applications. Based on the MVA algorithm, we deduce estimation formulas, which can be used to predict the overall speedup factor and the expected load at individual tiers. To evaluate our approach, we conduct a case study on a real world persistent identifier system.

Persistent identifiers are well acknowledged for providing an abstraction for addresses of research datasets. However, due to the explosive growth of research datasets the view onto the concept of persistent identification moves towards a much more fundamental component for research data management. The ability of attaching semantic information into persistent identifier records, in principle enables the realization of a virtual global research data network by means of persistent identifiers. However, the increased importance of persistent identifiers has at the same time led to a steadily increasing load at persistent identifier systems. Therefore, the focus of this paper is to propose a high performance persistent identifier management protocol. In contrast to the DNS system, persistent identifier systems are usually subjected to bulky registration requests originating from individual research data repositories. Thus, the fundamental approach in this work is to implement a bulk registration operation into persistent identifier systems. Therefore, in this work we provide an extended version of the established Handle protocol equipped with a bulk registration operation. Moreover, we provide a specification and an efficient data model for such a bulk registration operation. Finally, by means of a comprehensive evaluation, we show the profound speedup achieved by our extended version of the Handle System. This is also highly relevant for various other persistent identifier systems, which are based on the Handle System.

In current cloud computing systems, when leveraging virtualization technology, the customer’s requested data computing or storing service is accommodated by a set of communicated virtual machines (VM) in a scalable and elastic manner. These VMs are placed in one or more server nodes according to the node capacities or failure probabilities. The VM placement availability refers to the probability that at least one set of all customer’s requested VMs operates during the requested lifetime. In this paper, we first study the problem of placing at most$H$groups of$k$requested VMs on a minimum number of nodes, such that the VM placement availability is no less than$\delta$, and that the specified communication delay and connection availability for each VM pair under the same placement group are not violated. We consider this problem with and without Shared-Risk Node Group (SRNG) failures, and prove this problem is NP-hard in both cases. We subsequently propose an exact Integer Nonlinear Program (INLP) and an efficient heuristic to solve this problem. We conduct simulations to compare the proposed algorithms with two existing heuristics in terms of performance. Finally, we study the related reliable routing problem of establishing a connection over at most$w$link-disjoint paths from a source to a destination, such that the connection availability requirement is satisfied and each path delay is no more than a given value. We devise an exact algorithm and two heuristics to solve this NP-hard problem, and evaluate them via simulations.

Live Virtual Machine (VM) migration across data centers is an important support technology which will bring benefits to both cloud providers and users. At present, VM storage migration becomes the bottleneck of this technology as it explores only the static data feature of image files without much consideration on data semantics. In this paper, we propose a new space-efficient VM image structure-three-layer structure. According to different functions and features, the data of a VM are separated into an Operating System (OS) layer, a Working Environment (WE) layer, and a User Data (UD) layer. On basis of three-layer image structure, we design a novel VM storage migration system-LayerMover. It tries to improve the migration performance through optimizing current data deduplication methods. To evaluate the potential performance gains of LayerMover, we explore the similarity features of WE layer data. Our experimental results show that the similarity ratio between different images can reach 70%. Our experiments about the VM storage migration performance of LayerMover show that it can be 27% faster than current migration methods for twolayer image structures under our experimental environments. When migrating a VM in a second round or migrating a batch of VMs, LayerMover will see more performance improvement.

The immense research dataset growth requires new strategies and concepts for an appropriate handling at the corresponding research data repositories. This is especially true for the concept of Persistent Identifiers (PIDs), which is designed to provide a persistent identification layer on top of the address-based resource retrieval methodology of the current Internet. For research datasets, which are referenced, further processed and transferred, such a persistent identification concept is highly crucial for ensuring a long-term scientific exchange. Often, each individual research dataset stored in a research data repository, is registered at a particular PID registration agency in order to be assigned a globally unique PID. However, for the explosive growth of research datasets this concept of registering each individual research dataset is in terms of the performance highly inappropriate. Therefore, the focus of this work is on a concept for enabling a high-performance persistent identification of research datasets. Recent research data repositories often are equipped with a built-in naming component for assigning immutable and internally unique identifiers for their incoming research datasets. Thus, the core idea in this work is to enable these internal identifiers to be directly resolvable at the well-known global PID resolution systems. This work will therefore provide insight into the implementation of this idea into the well-known Handle System. Finally, in this work, we will provide an experimental evaluation of our proposed concept.

Summary Enterprises today are dealing with the massive size of data, which have been explosively increasing. The key requirements to address this challenge are to extract, analyze, and process data in a timely manner. Clustering is an essential data mining tool that plays an important role for analyzing big data. However, large‐scale data clustering has become a challenging task because of the large amount of information that emerges from technological progress in many areas, including finance and business informatics. Accordingly, researchers have dealt with parallel clustering algorithms using parallel programming models to address this issue. MapReduce is one of the most famous frameworks, and it has attracted great attention because of its flexibility, ease of programming, and fault tolerance. However, the framework has evident performance limitations, especially for iterative programs. This study will first review the proposed iterative frameworks that extended MapReduce to support iterative algorithms. We summarize these techniques, discuss their uniqueness and limitations, and explain how they address the challenging issues of iterative programs. We also perform an in‐depth review to understand the problems and the solving techniques for parallel clustering algorithms. Hence, we believe that no well‐rounded review provides a significant comparison among parallel clustering algorithms using MapReduce. This work aims to serve as a stepping stone for researchers who are studying big data clustering algorithms. Copyright © 2015 John Wiley & Sons, Ltd.

It is cost-efficient for a tenant with a limited budget to establish a virtual MapReduce cluster by renting multiple virtual private servers (VPSs) from a VPS provider. To provide an appropriate scheduling scheme for this type of computing environment, we propose in this paper a hybrid job-driven scheduling scheme (JoSS for short) from a tenant's perspective. JoSS provides not only job-level scheduling, but also map-task level scheduling and reduce-task level scheduling. JoSS classifies MapReduce jobs based on job scale and job type and designs an appropriate scheduling policy to schedule each class of jobs. The goal is to improve data locality for both map tasks and reduce tasks, avoid job starvation, and improve job execution performance. Two variations of JoSS are further introduced to separately achieve a better map-data locality and a faster task assignment. We conduct extensive experiments to evaluate and compare the two variations with current scheduling algorithms supported by Hadoop. The results show that the two variations outperform the other tested algorithms in terms of map-data locality, reduce-data locality, and network overhead without incurring significant overhead. In addition, the two variations are separately suitable for different MapReduce-workload scenarios and provide the best job performance among all tested algorithms.

Virtualized Cloud platforms have become increasingly common and the number of online services hosted on these platforms is also increasing rapidly. A key problem faced by providers in managing these services is detecting the performance anomalies and adjusting resources accordingly. As online services generate a very large amount of monitored data in the form of time series, it becomes very difficult to process this complex data by traditional approaches. In this work, we present a novel distributed parallel approach for performance anomaly detection. We build upon Holt-Winters forecasting for automatic aberrant behavior detection in time series. First, we extend the technique to work with MapReduce paradigm. Next, we correlate the anomalous metrics with the target Service Level Objective (SLO) in order to locate the suspicious metrics. We implemented and evaluated our approach on a production Cloud encompassing IaaS and PaaS service models. Experimental results confirm that our approach is efficient and effective in capturing the metrics causing performance anomalies in large time series datasets.

Renting a set of virtual private servers (VPSs for short) from a VPS provider to establish a virtual MapReduce cluster is cost-efficient for a company/organization. To shorten job turnaround time and keep data locality as high as possible in this type of environment, this paper proposes a Best-Fit Task Scheduling scheme (BFTS for short) from a tenant's perspective. BFTS schedules each map task to a VPS that can finish the task earlier than the other VPSs by predicting and comparing the time required by every VPS to retrieve the map-input data, execute the map task, and become idle in an online manner. Furthermore, BFTS schedules each reduce task to a VPS that is close to most VPSs that execute the related map tasks. We conduct extensive experiments to compare BFTS with several scheduling algorithms employed by Hadoop. The experimental results show that BFTS is better than the other tested algorithms in terms of map-data locality, reduce-data locality, and job turnaround time. The overhead incurred by BFTS is also evaluated, which is inevitable but acceptable compared with the other algorithms.

Virtualization technologies of Infrastructure-as-a- Service enable the live migration of running Virtual Machines (VMs) to achieve load balancing, fault-tolerance and hardware consolidation in data centers. However, the downtime/service unavailability due to live migration may be substantial with relevance to the customers' expectations on responsiveness, as the latter are declared in established Service Level Agreements (SLAs). Moreover, it may cause significant (potentially exponential) SLA violation penalties to its associated higher- level domains (Platform-as-a-Service and Software-as-a-Service). Therefore, VM live migration should be managed carefully. In this paper, we present the OpenStack version of the Generic SLA Manager, alongside its strategies for VM selection and allocation during live migration of VMs. We simulate a use case where IaaS (OpenStack-SLAM) and PaaS (OpenShift) are combined, and assess performance and efficiency of the aforementioned VM placement strategies, when a multi-domain SLA pricing & penalty model is involved. We find that our proposal is efficient in managing trade-offs between the operational objectives of service providers (including financial considerations) and the customers' expected QoS requirements.

Software defined networking (SDN) and OpenFlow as the outcome of recent research and development efforts provided unprecedented access into the forwarding plane of networking elements. This is achieved by decoupling the network control out of the forwarding devices. This separation paves the way for a more flexible and innovative networking. While SDN concept and OpenFlow find their ways into commercial deployments, performance evaluation of the SDN concept and its scalability, delay bounds, buffer sizing and similar performance metrics are not investigated in recent researches. In spite of usage of benchmark tools (like OFlops and Cbench), simulation studies and very few analytical models, there is a lack of analytical models to express the boundary condition of SDN deployment. In this work we present a model based on network calculus theory to describe the functionality of an SDN switch and controller. To the best of our knowledge, this is for the first time that network calculus framework is utilized to model the behavior of an SDN switch in terms of delay and queue length boundaries and the analysis of the buffer length of SDN controller and SDN switch. The presented model can be used for network designers and architects to get a quick view of the overall SDN network deployment performance and buffer sizing of SDN switches and controllers.

Cloud computing effectively implements the vision of utility computing by employing a pay-as-you-go cost model and allowing on-demand (re-)leasing of IT resources. Small or medium-sized Infrastructure-as-a-Service providers, however, find it challenging to satisfy all requests immediately due to their limited resource capacity. In that situation, both providers and customers may benefit greatly from advanced reservation of virtual resources, i.e. virtual machines. In our work, we assume SLA-based resource requests and introduce an advanced reservation methodology during SLA negotiation by using computational geometry. Thereby, we are able to verify, record and manage the infrastructure resources efficiently. Based on that model, service providers can easily verify the available capacity for satisfying the customer's Quality-of-Service requirements. Furthermore, we introduce flexible alternative counter-offers, when the service provider lacks resources. Therefore, our mechanism increases the utilization of the resources and attempts to satisfy as many customers as possible.

In modern service economies, service provisioning needs to be regulated by complex SLA hierarchies among providers of heterogeneous services, defined at the business, software, and infrastructure layers. Starting from the SLA Management framework defined in the SLA@SOI EU FP7 Integrated Project, we focus on the relationship between establishment and monitoring of such SLAs, showing how the two processes become tightly interleaved in order to provide meaningful mechanisms for SLA management. We first describe the process for SLA establishment adopted within the framework; then,we propose an architecture for monitoring SLAs, which satisfies the two main requirements introduced by SLA establishment: the availability of historical data for evaluating SLA offers and the assessment of the capability to monitor the terms in a SLA offer.

This paper addresses nonclairvoyant and non-preemptive online job scheduling in Grids. In the applied basic model, the grid system consists of a large number of identical processors that are divided into several machines. Jobs are independent, they have a fixed degree of parallelism, and they are submitted over time. Further, a job can only be executed on the processors belonging to the same machine. It is our goal to minimize the total makespan. We show that the performance of Garey and Graham's list scheduling algorithm is significantly worse in grids than in multiprocessors. Then we present a Grid scheduling algorithm that guarantees a competitive factor of 5. This algorithm can be implemented using a "job stealing" approach and may be well suited to serve as a starting point for Grid scheduling algorithms in real systems.

This paper presents QLSE (QoS-constrained list scheduling heuristics), a quality of service-based launch time scheduling algorithm for wide area grids. QLSE considers applications described by a task interaction graph (TIG) whose nodes and edges are labeled according to the Quality of Service requirements of the application. The high values obtained in the performance evaluation for both the tasks communication and computation throughput demonstrates the applicability of the proposed approach.

One of the key requirements for grid infrastructures is the ability to share resources with nontrivial qualities of service. However, resource management in a decentralized infrastructure is a complex task as it has to cope with different policies and objectives of the different resource providers and the resource users. Recent research indicates that agreement-based resource management will solve many of these problems as it supports the reliable interaction between different providers and users. Here, negotiation is needed to create such bi-lateral agreements between grid parties. Such negotiation processes should be automated with no or minimal human interaction, considering the potential scale of grid systems and the amount of necessary transactions. Therefore, strategic negotiation models play an important role. In this paper, a negotiation model and learning-based negotiation strategies are proposed and examined. Simulations have been conducted to evaluate the presented system. The results demonstrate that the proposed negotiation model and the learning based negotiation strategies are suitable and effective for grid environments.

Knowledge about the workload is an important aspect for scheduling of resources as parallel computers or grid components. As the scheduling quality highly depends on the characteristics of the workload running on such resources, a representative workload model is significant for performance evaluation. Previous approaches on workload modelling mainly focused on methods that use statistical distributions to fit the overall workload characteristics. Therefore, the individual association and correlation to users or groups are usually lost. However, job scheduling for single parallel installations as well as for grid systems started to focus more on the quality of service for specific-user groups. Here, detailed knowledge of the individual user characteristic and preference is necessary for developing appropriate scheduling strategies. In the absence of a large information base of actual workloads, the adequate modelling of submission behaviors is sought. In this paper, we propose a new workload model, called MUGM (mixed user group model), which maintains the characteristics of individual user groups. The MUGM method has been further evaluated by simulations and shown to yield good results.

This paper addresses the potential benefit of sharing jobs between independent sites in a grid computing environment. Also the aspect of parallel multi-site job execution on different sites is discussed. To this end, various scheduling algorithms have been simulated for several machine configurations with different workloads which have been derived from real traces. The results showed that a significant improvement in terms of a smaller average response time is achievable. The usage of multi-site applications can additionally improve the results as long as the increase of the execution time due to communication overhead is limited to about 25%.