Performance Of Virtual Machines Research Articles

IConSnap: An Incremental Continuous Snapshots System for Virtual Machines

The reliability of data and services hosted on a virtual machine (VM) is a top concern in cloud environments. The Continuous Snapshots can reduce the data loss in case of failures and thus is prevailing for protecting long-running systems. However, existing methods suffer from long VM downtime, long snapshot interval and significant performance loss. In this paper, we present iConsnap, a system designed to take fine-grained continuous snapshots of virtual machines without compromising VM performance. First, iConsnap adopts the copy-on-write (COW) mechanism to save the memory pages on-demand, and thus decreases the downtime to about 200 milliseconds. Second, we extend the idea of COW and propose a lazily incremental approach to save the delta data between two successive snapshots only once, thereby reducing the snapshot duration and snapshot data a lot. Third, we propose several scheduling mechanisms to mitigate the VM performance penalty issue. Last, we introduce a method combined of compression and time-aware multi-granularity reclamation strategy to reduce the storage costs without losing performance and availability. We implement iConsnap on QEMU/KVM and evaluate it through a set of experiments. The results show that iConsnap outperforms existing approaches in terms of downtime, duration, storage costs and VM performance.

Predictive Auto-Scaling of Multi-Tier Applications Using Performance Varying Cloud Resources

The performance of the same type of cloud resources, such as virtual machines (VMs), varies over time mainly due to hardware heterogeneity, resource contention among co-located VMs, and virtualization overhead. The performance variation can be significant, introducing challenges to learn workload-specific resource provisioning policies to automatically scale the cloud-hosted applications to maintain the desired response time. Moreover, auto-scaling multi-tier applications using minimal resources is even more challenging because bottlenecks may occur on multiple tiers concurrently. In this paper, we address the problem of using performance varying VMs for gracefully auto-scaling a multi-tier application using minimal resources to handle dynamically increasing workloads and satisfy the response time requirements. The proposed system uses a supervised learning method to identify the appropriate resources provisioning for multi-tier applications based on the prediction of the application response time and the request arrival rate. The supervised learning method learns a state transition configuration map which encodes a resource allocation states invariant to the underlying VMs performance variations. This configuration map helps to use performance varying resources in predictive autoscaling method. Our experimental evaluation using a real-world multi-tier web application hosted on a public cloud shows an improved application performance with minimal resources compared to conventional predictive auto-scaling methods.

CPU Overclocking: A Performance Assessment of Air, Cold Plates, and Two-Phase Immersion Cooling

Computing capacity has always been on the upward climb due to the constant technological improvements in semiconductor manufacturing and packaging industry. This growth in computing capability is usually accompanied by a steep rise in heat flux density associated with the electronic component (CPUs or GPUs, for example). High-performance computing (HPC) data centers often employ several of these high-performance devices for crunching their enormous artificial intelligence (AI) or scientific computing workloads. State-of-the-art air cooling technologies throttle after a certain heat flux level and would require bigger heat sinks driving enormous airflow through it, which may not be desirable from a data center operation standpoint. Hence, a lookout for advanced thermal management techniques is quite imperative. In this article, a commercially available intel overclockable CPU i9-9900k was exercised with high performance workloads while employing and evaluating three different cooling technologies namely air-cooled, cold plates, and two-phase (2P) immersion. The high heat carrying capacity associated with cold plates and 2P immersion techniques outperformed the air-cooled solution by constantly yielding higher CPU clock rates up to 41% utilizing cold plates and 51% employing 2P immersion. The performance of cold plates and 2P immersion was evaluated at different functional points (different coolant operating temperatures and 2P coolant types for example) to understand how the technologies would respond in a legacy or modern data center operation. It was again observed that both 2P immersion technology and cold plates provided the least thermal resistance (as low as 0.247 °C/W) path to heat transfer and therefore provided higher computational performance and efficiency compared to air cooling. Further improvement in performance is not limited by cooling but by packaging constraints. Advanced packaging techniques coupled with minimizing the interfacial resistance can reap enhanced performance and energy efficiency gains using these modes of cooling. This article demonstrates the potential increase in virtual machine (VM) performance that can be attained using cold plates or 2P immersion. Authors recommend adoption of liquid cooling in early chip design and architecture phases. Finally, the article provides a generic overclocking methodology that the industry can use to qualify postturbo server class chip performance.

Software-defined networks for resource allocation in cloud computing: A survey

Cloud computing has a shared set of resources, including physical servers, networks, storage, and user applications. Resource allocation is a critical issue for cloud computing, especially in Infrastructure-as-a-Service (IaaS). The decision-making process in the cloud computing network is non-trivial as it is handled by switches and routers. Moreover, the network concept drifts resulting from changing user demands are among the problems affecting cloud computing. The cloud data center needs agile and elastic network control functions with control of computing resources to ensure proper virtual machine (VM) operations, traffic performance, and energy conservation. Software-Defined Network (SDN) proffers new opportunities to blueprint resource management to handle cloud services allocation while dynamically updating traffic requirements of running VMs. The inclusion of an SDN for managing the infrastructure in a cloud data center better empowers cloud computing, making it easier to allocate resources. In this survey, we discuss and survey resource allocation in cloud computing based on SDN. It is noted that various related studies did not contain all the required requirements. This study is intended to enhance resource allocation mechanisms that involve both cloud computing and SDN domains. Consequently, we analyze resource allocation mechanisms utilized by various researchers; we categorize and evaluate them based on the measured parameters and the problems presented. This survey also contributes to a better understanding of the core of current research that will allow researchers to obtain further information about the possible cloud computing strategies relevant to IaaS resource allocation.

On the Performance of Cloud-based Virtual Machines

On the Performance of Cloud-based Virtual Machines

Benchmarking and Performance Evaluations on Various Configurations of Virtual Machine and Containers for Cloud-Based Scientific Workloads

Cloud computing manages system resources such as processing, storage, and networking by providing users with multiple virtual machines (VMs) as needed. It is one of the rapidly growing fields that come with huge computational power for scientific workloads. Currently, the scientific community is ready to work over the cloud as it is considered as a resource-rich paradigm. The traditional way of executing scientific workloads on cloud computing is by using virtual machines. However, the latest emerging concept of containerization is growing more rapidly and gained popularity because of its unique features. Containers are treated as lightweight as compared to virtual machines in cloud computing. In this regard, a few VMs/containers-associated problems of performance and throughput are encountered because of middleware technologies such as virtualization or containerization. In this paper, we introduce the configurations of VMs and containers for cloud-based scientific workloads in order to utilize the technologies to solve scientific problems and handle their workloads. This paper also tackles throughput and efficiency problems related to VMs and containers in the cloud environment and explores efficient resource provisioning by combining four unique methods: hyperthreading (HT), vCPU cores selection, vCPU affinity, and isolation of vCPUs. The HEPSCPEC06 benchmark suite is used to evaluate the throughput and efficiency of VMs and containers. The proposed solution is to implement four basic techniques to reduce the effect of virtualization and containerization. Additionally, these techniques are used to make virtual machines and containers more effective and powerful for scientific workloads. The results show that allowing hyperthreading, isolation of CPU cores, proper numbering, and allocation of vCPU cores can improve the throughput and performance of virtual machines and containers.

Proactive load balancing fault tolerance algorithm in cloud computing

AbstractAs the demand for cloud computing infrastructure increased, availability and reliability have become more important as they are major features in real‐time computing systems. The chances of failure in a cloud computing system increase compared to other computing systems. To achieve high reliability and availability for the cloud computing infrastructure fault tolerance is one of the most essential modules. For a cloud computing infrastructure, a proactive fault tolerance algorithm with load balancing is proposed in this paper. The proposed methodology tolerates virtual machine CPU faults to achieve maximum cloud computing infrastructure reliability and availability. CPU faults can occur during virtual machine operation. The proposed model's main objective is to track changes in CPU utilization and make a decision once a high value of CPU utilization is identified. The decision is either to migrate the defective virtual machine to another “destination host”, or to manage loads on the destination host so that the faulty virtual machine can be migrated. A load is calculated for the destination host; if the destination host is overloaded, the subroutine for load balancing is launched. Load balancing is based on the technique of choosing a virtual machine.

Small‐signal analysis of multiple virtual synchronous machines to enhance frequency stability of grid‐connected high renewables

The virtual synchronous machine technology is considered as an important technique to effectively control the shortcomings of renewable energy-based power electronics interfaces, providing backup inertia and regulating grid stability. Conventionally, the virtual synchronous machine with a large capacity is responsible for controlling the entire grid stability against renewable penetration. It is usually operated as a centralised control system. But what if virtual synchronous machines with small capacities are independently operated by their additional droop control schemes, and will they present better performance than the single virtual synchronous machine? This study proposes the multiple virtual synchronous machine system with different active power-frequency (P-f) droop characteristics to improve inertia support regarding frequency stability improvement. The comprehensive small-signal modelling of the multiple virtual synchronous machine unit is designed to include the additional P-f droop characteristics. Then, the dynamic characteristics (steady-state and transient responses) and static stability of the multiple virtual synchronous machines are compared with the single virtual synchronous machine at the same rated capacity in both eigenvalue/sensitivity-domain and time-domain analysis. The obtained results reveal that the system with the presence of several virtual synchronous machines is more stable than the system with single virtual synchronous machine, maintaining stable and secure system operation during the contingency.

Migration-Based Load Balance of Virtual Machine Servers in Cloud Computing by Load Prediction Using Genetic-Based Methods

This paper presents a two-stage genetic mechanism for the migration-based load balance of virtual machine hosts (VMHs) in cloud computing. Previous methods usually assume this issue as a job-assignment optimization problem and only consider the current VMHs' loads; however, without considering loads of VMHs after balancing, these methods can only gain limited effectiveness in real environments. In this study, two genetic-based methods are integrated and presented. First, performance models of virtual machines (VMs) are extracted from their creating parameters and corresponding performance measured in a cloud computing environment. The gene expression programming (GEP) is applied for generating symbolic regression models that describe the performance of VMs and are used for predicting loads of VMHs after load-balance. Secondly, with the VMH loads estimated by GEP, the genetic algorithm considers the current and the future loads of VMHs and decides an optimal VM-VMH assignment for migrating VMs and performing load-balance. The performance of the proposed methods is evaluated in a real cloud-computing environment, Jnet, wherein these methods are implemented as a centralized load balancing mechanism. The experimental results show that our method outperforms previous methods, such as heuristics and statistics regression.

Benchmarking Computational Performance of Virtual Machines Within the Secure Unified Research Environment Versus Physical Computers

IntroductionThe Secure Unified Research Environment (SURE) is a high-powered computing environment located within Sax Institute (Sydney, Australia). SURE was established through the financial support of the Australian Government National Collaborative Research Infrastructure Strategy (NCRIS) as part of the Population Health Research Network (PHRN). SURE is approved by the Australian Government as the only secure platform for analysing unit record level sensitive health and other Australian Government data, providing computational resources and secure infrastructure in a form of virtual machines (VMs) accessible by approved researchers in Australia and overseas.
 Objectives and ApproachWe aim to compare computational performance of SURE VMs of different configurations with the performance of physical computers by running a series of standardised computational tasks involving different numbers of central processingunit (CPU) cores available on each computer. The approach utilised the benchmark test maintained by the H2O.ai group (https://h2oai.github.io/db-benchmark/). The results were measured over the datasets of different sizes, ranging from 500MB to 50GB in Random Access Memory (RAM).
 ResultsOur benchmarking outcomes have revealed that computational efficiency of physical computers uniformly outperform the efficiency of the current standard SURE VM configuration offerings, sometimes demonstrating a nearly double performance. For the range of typical analytical tasks assessed, computational performance greatly benefits from extending the number of computational cores available on a machine.
 Conclusion / ImplicationsSURE is a highly valuable tool enabling research and collaborations involving confidential population-based data. The shortage of RAM and CPU cores can be a major bottleneck even for moderately large datasets. VMs currently offered by SURE yet fall short of reaching computational performance of physical desktop computers. The results are to guide the funders and providers of secure remote access data laboratories responsible for providing Research Infrastructure as a Service (IaaS) tailored to meet the needs of participating research groups.

Hybrid Improved Pre Copy and Post Copy Approach for Virtual Machine Migration in Cloud

With the increasing of rapid demand for cloud resources, load on the cloud server is increased drastically in the past few decades. Live virtual machine (VM) relocation is an important approach for the dynamic scheduling of resources in the cloud environment. The solution for dealing with the dynamic load that transfers the virtual machine from one server to another is virtual machine migration. The cloud provider uses three virtual machine migration techniques such as stop and copy, pre-copy, and post-copy. The pre-copy approach is the most popular approach among all migration approaches but it is not suitable for the write intensive application is running on the virtual machine. This article suggested a hybrid virtual machine solution that would minimize the overall migration time, and downtime. And also there is an increase in the virtual machine's performance. To test the efficiency of the suggested solution, the CloudSim simulation tool is used. Compared to the existing method, the findings of the experiments indicate that the proposed method produces better results.

Exploring Impact of Profile Data on Code Quality in the HotSpot JVM

Managed language virtual machines (VM) rely on dynamic or just-in-time (JIT) compilation to generate optimized native code at run-time to deliver high execution performance. Many VMs and JIT compilers collect profile data at run-time to enable profile-guided optimizations (PGO) that customize the generated native code to different program inputs. PGOs are generally considered integral for VMs to produce high-quality and performant native code. In this work, we study and quantify the performance benefits of PGOs, understand the importance of profiling data quantity and quality/accuracy to effectively guide PGOs, and assess the impact of individual PGOs on VM performance. The insights obtained from this work can be used to understand the current state of PGOs, develop strategies to more efficiently balance the cost and exploit the potential of PGOs, and explore the implications of and challenges for the alternative ahead-of-time (AOT) compilation model used by VMs.

NetAP: Adaptive Polling Technique for Network Packet Processing in Virtualized Environments

In cloud systems, computing resources, such as the CPU, memory, network, and storage devices, are virtualized and shared by multiple users. In recent decades, methods to virtualize these resources efficiently have been intensively studied. Nevertheless, the current virtualization techniques cannot achieve effective I/O virtualization when packets are transferred between a virtual machine and a host system. For example, VirtIO, which is a network device driver for KVM-based virtualization, adopts an interrupt-based packet-delivery mechanism, and incurs frequent switch overheads between the virtual machine and the host system. Therefore, VirtIO wastes valuable CPU resources and decreases network performance. To address this limitation, this paper proposes an adaptive polling-based network I/O processing technique, called NetAP, for virtualized environments. NetAP processes network requests via a periodical polling-based mechanism. For this purpose, NetAP adopts the golden-section search algorithm to determine the near-optimal polling interval for various workloads with different characteristics. We implement NetAP in a Linux kernel and evaluated it with up to six virtual machines. The evaluation results show that NetAP can improve the network performance of virtual machines by up to 31.16%, while only using 32.92% of the host CPU time used by VirtIO for packet processing.

Memory-aware resource management algorithm for low-energy cloud data centers

The continuous advancement of cloud computing technology has driven the vigorous development of cloud data centers. This manifests itself not only in increasing numbers, but also in rapid expansion scale. At the same time, the problems of high energy consumption, high cost and high carbon emissions began to stand out. These factors have become a bottleneck restricting the further development of cloud computing technology. As a result of the application of virtualization technology, mature cloud service providers use virtual machines instead of physical machines to provide users with computing, storage and other services. Therefore, the scheduling algorithm of cloud tasks and virtual machines has been widely studied by academia as a core problem. However, most of the current cloud data center resource management algorithms take CPU utilization as the main consideration, and increase the CPU utilization through virtual machine integration to reduce the energy consumption of cloud data centers. However, these algorithms will cause waste and low utilization of other resources in the cloud data center due to excessive consideration of CPU utilization. This paper systematically analyzes the mapping relationship between cloud tasks, virtual machines and physical machines. At the same time, the performance of cloud tasks, virtual machines, and physical machines is modeled in the cloud data center. By comprehensively considering the CPU and memory characteristics of cloud tasks, the memory priority cloud task mapping rule is established. Based on the rule, the memory-aware resource management algorithm for low-energy cloud data centers (MALE) is proposed. The algorithm maps cloud tasks and deploys virtual machines according to the memory requirements of the cloud tasks, thereby achieving the goal of simultaneously reducing the total fee of cloud users and the energy consumption of cloud data centers. Finally, the algorithm is compared with the other three algorithms. The experimental simulation results show that the effect of the proposed algorithm in this paper is significantly better than the comparison algorithm for the total fee of cloud users and the energy consumption of the cloud data center.

An Observable Network Route Support on Interpretation of Cloud Computing

The Commercial cloud computing is appropriate conventional and funding agencies beyond prototyping, and initiated fund Production exercise. An important feature of any technical computing Program is moving production data, inward and outward. By means of the virtual machine performance and cost relatively glowing assumed, Network performance and cost is not. This article provides an authentication in the regions of Amazon Web Services, Microsoft Azure network and between Google clouds platforms, cloud both resources and major DTNs In research platform in the Pacific, including the Federation of OSG data cache Network backbone, cloud inside their own. This article contains both qualitative results of the analysis, as well as latency and throughput measuring. It also includes analysis of the cost of contribution Cloud Based on the network.

A Hybrid Cloud Testing System Based on Virtual Machines and Networks

Traditional software testing typically uses many physical resources to manually build various test environments, resulting in high resource costs and long test time due to limited resources, especially for small enterprises. Cloud computing can provide sufficient low-cost virtual resources to alleviate these problems through the virtualization of physical resources. However, the provision of various test environments and services for implementing software testing rapidly and conveniently based on cloud computing is challenging. This paper proposes a multilayer cloud testing model based on cloud computing and implements a hybrid cloud testing system based on virtual machines (VMs) and networks. This system realizes the automatic and rapid creation of test environments and the remote use of test tools and test services. We conduct experiments on this system and evaluate its applicability in terms of the VM provision time, VM performance and virtual network performance. The experimental results demonstrate that the performance of the VMs and virtual networks is satisfactory and that this system can improve the test efficiency and reduce test costs through rapid virtual resource provision and convenient test services.

Live Virtual Machine migration: A Performance Test Xen and KVM

Virtualization technology has many important features such as live virtual machine migration. In live virtual machine migration, a power on virtual machine is moved from one physical host to another. It has various benefits such as server consolidation, proactive failure, load balancing, energy saving and resource scheduling. Live virtual machine migration is very useful tool in cluster environment, administrators of data centers and in cloud environment. Live virtual machine migration is supported by hypervisors such as Xen, KVM, VMware etc. In this paper we discuss live virtual machine migration Pre-Copy approach which is a default approach in many hypervisors. We compare the performance of virtual machines which are made using Xen and KVM. We also compare performance when virtual machines are migrate using Xen and KVM in cloudreport simulator. In result we find that KVM performs better than Xen.

Energy-efficient and Dynamic Consolidation of Virtual Machines in OpenStack-based Private Cloud

Abstract Dynamic Virtual Machines (VMs) consolidation is an efficient technique for enhancing resource utilization and reducing energy consumption by the physical servers. Over-utilized host affects the degradation of resource utilization and VM performance. Determining when it is best to migrate VMs from an over-utilized host to an under-utilized host is the aim of dynamic Live VM Migration (LVM). It influences the resource utilization and Quality of Services (QoS) offered by cloud providers. In the OpenStack cloud, LVM is done manually, and it uses first-fit as a default migration technique, which is not efficient for resource utilization. The paper proposes a novel approach for the dynamic consolidation of VMs in the OpenStack cloud. The CPU utilization, RAM utilization, and the number of instances of each host are monitored over a period of time, and overload detection of a host is carried out using the SVM classification model. The Load balancing consolidation is performed based on the SVM classification result. Further, an energy-efficient consolidation achieved to optimize the energy of physical servers. We carry out the performance analysis of the proposed work in the OpenStack-based real-time testbed of five-node. The result reveals the fair distribution of load within the servers and the significant benefits of energy-efficient VM consolidation resulting up to 15% energy savings.

Test environment for verification of multi-processor interrupt system with virtualization support

Interrupt system is an important part of microprocessors. Interrupts are widely used for interaction with hardware and responding to stimuli. Modern microprocessor interrupt systems include hardware support of virtualization. Hardware support helps to increase the performance of virtual machines. However, including additional functionality may lead to potential errors. The paper presents an overview of approaches used for multi-core microprocessors interrupt system with virtualization support verification. Some definitions and characteristics of interrupt systems that needed to be taken into account in the process of verification are described. Stand-alone verification environment general scheme is presented. Universal Verification Methodology was applied to construct test system. To simplify development of checking module discrete-event with time accounting reference model was used. Sequences of primary requests and automatically generated secondary requests in the special modules named auto-handlers were used for test system behavior randomization. We describe some difficulties discovered in the verification process and corresponding solving methods. Generalized test algorithm stages are presented. Some other techniques for checking the correctness of interrupt system have been reviewed. In conclusion, we provide the case study of applying the suggested approaches for interrupt system verification of microprocessors with “Elbrus” and “SPARC-V9” architectures developed by MCST. The results and further plan of the test system development are presented.

A Resource Usage Intensity Aware Load Balancing Method for Virtual Machine Migration in Cloud Datacenters

To provide robust infrastructure as a service (IaaS), clouds currently perform load balancing by migrating virtual machines (VMs) from heavily loaded physical machines (PMs) to lightly loaded PMs. The unique features of clouds pose formidable challenges to achieving effective and efficient load balancing. First, VMs in clouds use different resources (e.g., CPU, bandwidth, memory) to serve a variety of services (e.g., high performance computing, web services, file services), resulting in different overutilized resources in different PMs. Also, the overutilized resources in a PM may vary over time due to the time-varying heterogeneous service requests. Second, there is intensive network communication between VMs. However, previous load balancing methods statically assign equal or predefined weights to different resources, which lead to degraded performance in terms of speed and cost to achieve load balance. Also, they do not strive to minimize the VM communications between PMs. We propose a Resource Intensity Aware Load balancing method (RIAL). For each PM, RIAL dynamically assigns different weights to different resources according to their usage intensity in the PM, which significantly reduces the time and cost to achieve load balance and avoids future load imbalance. It also tries to keep frequently communicating VMs in the same PM to reduce bandwidth cost, and migrates VMs to PMs with minimum VM performance degradation. We also propose an extended version of RIAL with three additional algorithms. First, it optimally determines the weights for considering communication cost and performance degradation due to VM migrations. Second, it has a more strict migration triggering algorithm to avoid unnecessary migrations while still satisfying Service Level Objects (SLOs). Third, it conducts destination PM selection in a decentralized manner to improve scalability. Our extensive trace-driven simulation results and real-world experimental results show the superior performance of RIAL compared to other load balancing methods.