Alexander Yu. Drozdov

Online Bi-Objective Scheduling for IaaS Clouds Ensuring Quality of Service

This paper focuses on a bi-objective experimental evaluation of online scheduling in the Infrastructure as a Service model of Cloud computing regarding income and power consumption objectives. In this model, customers have the choice between different service levels. Each service level is associated with a price per unit of job execution time, and a slack factor that determines the maximal time span to deliver the requested amount of computing resources. The system, via the scheduling algorithms, is responsible to guarantee the corresponding quality of service for all accepted jobs. Since we do not consider any optimistic scheduling approach, a job cannot be accepted if its service guarantee will not be observed assuming that all accepted jobs receive the requested resources. In this article, we analyze several scheduling algorithms with different cloud configurations and workloads, considering the maximization of the provider income and minimization of the total power consumption of a schedule. We distinguish algorithms depending on the type and amount of information they require: knowledge free, energy-aware, and speed-aware. First, to provide effective guidance in choosing a good strategy, we present a joint analysis of two conflicting goals based on the degradation in performance. The study addresses the behavior of each strategy under each metric. We assess the performance of different scheduling algorithms by determining a set of non-dominated solutions that approximate the Pareto optimal set. We use a set coverage metric to compare the scheduling algorithms in terms of Pareto dominance. We claim that a rather simple scheduling approach can provide the best energy and income trade-offs. This scheduling algorithm performs well in different scenarios with a variety of workloads and cloud configurations.

Biobjective VoIP Service Management in Cloud Infrastructure

Voice over Internet Protocol (VoIP) allows communication of voice and/or data over the internet in less expensive and reliable manner than traditional ISDN systems. This solution typically allows flexible interconnection between organization and companies on any domains. Cloud VoIP solutions can offer even cheaper and scalable service when virtualized telephone infrastructure is used in the most efficient way. Scheduling and load balancing algorithms are fundamental parts of this approach. Unfortunately, VoIP scheduling techniques do not take into account uncertainty in dynamic and unpredictable cloud environments. In this paper, we formulate the problem of scheduling of VoIP services in distributed cloud environments and propose a new model for biobjective optimization. We consider the special case of the on-line nonclairvoyant dynamic bin-packing problem and discuss solutions for provider cost and quality of service optimization. We propose twenty call allocation strategies and evaluate their performance by comprehensive simulation analysis on real workload considering six months of the MIXvoip company service.

PHEFT: Pessimistic Image Processing Workflow Scheduling for DSP Clusters

We address image processing workflow scheduling problems on a multicore digital signal processor cluster. We present an experimental study of scheduling strategies that include task labeling, prioritization, resource selection, and digital signal processor scheduling. We apply these strategies in the context of executing the Ligo and Montage applications. To provide effective guidance in choosing a good strategy, we present a joint analysis of three conflicting goals based on performance degradation. A case study is given, and experimental results demonstrate that a pessimistic scheduling approach provides the best optimization criteria trade-offs. The Pessimistic Heterogeneous Earliest Finish Time scheduling algorithm performs well in different scenarios with a variety of workloads and cluster configurations.

A VHDL-Based Modeling of Network Interface Card Buffers: Design and Teaching Methodology

The design of High Performance Computing (HPC) relies to a large extent on simulations to optimize components of such complex systems. A key hardware component of the interconnection network in HPC environments is the Network Interface Card (NIC). In spite of the popularity of simulation-based approaches in the computer architecture domain, few authors have focused on simulators design methodologies. In this paper, we describe the stages of implementing a simulation model to solve a real problem—modeling NIC buffer. We present a general methodology for helping users to build Hardware Description Language (HDL)/SystemC models targeted to fulfil features such as performance evaluation of compute nodes. The developed VHDL model allows reproducibility and can be used as a tool in the area of HPC education.

WA-RRNS: Reliable Data Storage System Based on Multi-cloud

Cloud technologies are widely used for storage services. However, the single cloud cannot ensure the reliability of data. To solve the security issue, we present a multi-cloud based storage architecture called WA-RRNS that combines weighted access scheme and threshold secret sharing redundant residue number system with multiple failure detection/recovery mechanisms and homomorphic ciphers. For better tradeoffs between security and performance, WA-RRNS uses parameters to adjust redundancy, encryption-decryption speed, and data loss probability. Theoretical and experimental analysis with real data shows that our approach provides a secure way to mitigate the uncertainty of the use of untrusted and not reliable cloud storage.

Bi-objective Heterogeneous Consolidation in Cloud Computing

In this paper, we address the problem of power-aware Virtual Machines (VMs) consolidation considering resource contention. Deployment of VMs can greatly influence host performance, especially, if they compete for resources on insufficient hardware. Performance can be drastically reduced and energy consumption increased. We focus on a bi-objective experimental evaluation of scheduling strategies for CPU and memory intensive jobs regarding the quality of service (QoS) and energy consumption objectives. We analyze energy consumption of the IBM System x3650 M4 server, with optimized performance for business-critical applications and cloud deployments built on IBM X-Architecture. We create power profiles for different types of applications and their combinations using SysBench benchmark. We evaluate algorithms with workload traces from Parallel Workloads and Grid Workload Archives and compare their non-dominated Pareto optimal solutions using set coverage and hyper volume metrics. Based on the presented case study, we show that our algorithms can provide the best energy and QoS trade-offs.

Robust cloud VoIP scheduling under VMs startup time delay uncertainty

In this paper, we address cloud VoIP service orchestration and scheduling to provide appropriate levels of quality of service to users, and performance to VoIP service providers. We consider voice quality affected by call processing, and cost contributed by billing hours for used VMs in a cloud. We believe that this biobjective focus is reasonable and representative for real installations and applications. We conduct comprehensive simulation of our calls load balancing strategies on real data and show that not all approaches provide suitable quality of service. We analyze eight on-line dynamic non-clairvoyant scheduling strategies with variations in VM startup time delays to deal with realistic VoIP cloud environments. We show that the proposed strategies outperform currently in use strategies in terms of quality of service and provider cost. The robustness of these strategies is also discussed.