KVM Virtual Machine Research Articles

Combining containers and virtual machines to enhance isolation and extend functionality on cloud computing

Virtualization technology is the underlying element of cloud computing. Traditionally, cloud computing has employed virtual machines to distribute available resources and provide isolated environments among users. Multiple virtual machines, with their own operating system and services, can be deployed and run simultaneously on the same physical machine on the cloud infrastructure. Recently, a more lightweight virtualization technology is being rapidly adopted and it is based on containers. A key difference between virtual machines and containers, is that containers share the same underlying operating system.In this work, we studied the combination of these two virtualization technologies by running containers on top of virtual machines. This synergy aims to enhance containers’ main drawback, which is isolation and, among others, to simplify the system management and upgrade, and to introduce the new functionalities of containerized applications to virtual machines. The benefits of this method have been recognized by big cloud companies, which have been employing this approach for years.With this paper, we aimed to present the advantages of running containers on virtual machines and to explore how various virtualization techniques and configurations affect the performance of this method. Although, in bibliography there are a few papers that used this method partially to conduct experiments, there is not any research which considers this method from our integral aspect of view. In this study, we ran Docker containers and evaluated their performances, not only on KVM and XEN virtual machines, but also we ran Linux containers on Windows Server. We adopted an empirical approach, to quantify the performance overhead introduced by the additional virtualization layer of virtual machines, by executing various benchmarks and deploying real world applications as use cases. Furthermore, we presented for first time, how isolation is applied on virtual machines and containers, we evaluated different operating systems optimized to host containers, mechanisms of storing persistent data and, last but not least, we experimentally quantified the power and energy consumption overhead of containers running on virtual machines.

Instance launch-time analysis of OpenStack virtualization technologies with control plane network errors

We analyzed the performance of a multi-node OpenStack cloud amid different types of controlled and self-induced network errors between controller and compute-nodes on the control plane network. These errors included limited bandwidth, delays and packet losses of varying severity. This study compares the effects of network errors on spawning times of batches of instances created using three different virtualization technologies supported by OpenStack, i.e., Docker containers, Linux containers and KVM virtual machines. We identified minimum/maximum thresholds for bandwidth, delay and packet-loss rates below/beyond which instances fail to launch. To the authors’ best knowledge, this is the first comparative measurement study of its kind on OpenStack. The results will be of particular interest to designers and administrators of distributed OpenStack deployments.

Performance overhead analysis of virtualisation on ARM

There is growing interest to replace traditional servers with low power systems. ARM has released new architecture with hardware virtualisation extension support. However, the performance of virtualisation on ARM-based low power platform is unclear. In this paper, a number of performance measurements on basic operations such as disk speed, CPU, memory and network throughput were done on a dual-core ARM Cortex-A7-based hardware platform with two popular open source hypervisor software, Xen and KVM. Basic benchmarks and three popular applications are measured to make a comparison between virtual machine and native machine. Our result shows that the average performance overhead of Xen and KVM virtual machine is between 3% and 4% when the host is lightly loaded, while the performance of three applications drops drastically with the workload stress increase in the system. The IO virtualisation should be optimised for the industrial usage of ARM-based virtualisation server.

Development of cloud services for patient-specific simulations of blood flows through aortic valves

The paper presents the development of cloud software services for patient-specific computational analysis of blood flows through the aortic valve on a private university cloud. The main focus is on the software service level at the top of the provided computational platform. Blood flow through the aortic valve was considered as a pilot application of the OpenStack cloud infrastructure. A modelling software environment based on ANSYS Fluent was developed as a software service (SaaS) for the numerical analysis of low flow, low pressure gradient aortic stenosis. Segmentation software services were designed to deal with the patient-specific issues of the computational analysis. User-friendly management tools were developed using Apache jclouds API to enhance the management of OpenStack cloud infrastructure and to increase the accessibility of the required software. The performance of the cloud infrastructure was assessed by testing CPU, memory bandwidth, disk I/O and the developed software service for medical computations. The performance measured on Xen hardware virtual machines, KVM virtual machines and Docker containers were compared with the performance obtained by using the native hardware.

AN INTERACTIVE WEB-BASED ANALYSIS FRAMEWORK FOR REMOTE SENSING CLOUD COMPUTING

Abstract. Spatiotemporal data, especially remote sensing data, are widely used in ecological, geographical, agriculture, and military research and applications. With the development of remote sensing technology, more and more remote sensing data are accumulated and stored in the cloud. An effective way for cloud users to access and analyse these massive spatiotemporal data in the web clients becomes an urgent issue. In this paper, we proposed a new scalable, interactive and web-based cloud computing solution for massive remote sensing data analysis. We build a spatiotemporal analysis platform to provide the end-user with a safe and convenient way to access massive remote sensing data stored in the cloud. The lightweight cloud storage system used to store public data and users’ private data is constructed based on open source distributed file system. In it, massive remote sensing data are stored as public data, while the intermediate and input data are stored as private data. The elastic, scalable, and flexible cloud computing environment is built using Docker, which is a technology of open-source lightweight cloud computing container in the Linux operating system. In the Docker container, open-source software such as IPython, NumPy, GDAL, and Grass GIS etc., are deployed. Users can write scripts in the IPython Notebook web page through the web browser to process data, and the scripts will be submitted to IPython kernel to be executed. By comparing the performance of remote sensing data analysis tasks executed in Docker container, KVM virtual machines and physical machines respectively, we can conclude that the cloud computing environment built by Docker makes the greatest use of the host system resources, and can handle more concurrent spatial-temporal computing tasks. Docker technology provides resource isolation mechanism in aspects of IO, CPU, and memory etc., which offers security guarantee when processing remote sensing data in the IPython Notebook. Users can write complex data processing code on the web directly, so they can design their own data processing algorithm.

On the performanc of KVM-based virtual routers

This paper presents an extensive experimental evaluation of the layer 3 packet forwarding performance of virtual software routers based on the Linux kernel and the KVM virtual machine. The impact of various tuning and configuration options on forwarding performance is evaluated, focussing on the mechanism used for moving data to and from virtual machines, the algorithm used for scheduling the virtual router tasks, the number of used CPU cores, and the router tasks affinities. The presented results show how to properly configure the virtual router components to improve forwarding performance and the benefits of using appropriate CPU schedulers. Furthermore, some advanced architectures based on virtual router aggregation are evaluated. The presented experiments show that architectures based on router aggregation can better exploit the available CPU cores to reach performance not far from the ones obtained by non-virtualised software routers.

Supporting High Performance Molecular Dynamics in Virtualized Clusters using IOMMU, SR-IOV, and GPUDirect

Cloud Infrastructure-as-a-Service paradigms have recently shown their utility for a vast array of computational problems, ranging from advanced web service architectures to high throughput computing. However, many scientific computing applications have been slow to adapt to virtualized cloud frameworks. This is due to performance impacts of virtualization technologies, coupled with the lack of advanced hardware support necessary for running many high performance scientific applications at scale. By using KVM virtual machines that leverage both Nvidia GPUs and InfiniBand, we show that molecular dynamics simulations with LAMMPS and HOOMD run at near-native speeds. This experiment also illustrates how virtualized environments can support the latest parallel computing paradigms, including both MPI+CUDA and new GPUDirect RDMA functionality. Specific findings show initial promise in scaling of such applications to larger production deployments targeting large scale computational workloads.

Performance overhead of KVM on Linux 3.9 on ARM cortex-a15

A number of simple performance measurements on network, CPU and disk speed were done on a dual ARM Cortex-A15 machine running Linux inside a KVM virtual machine that uses virtio disk and networking. Unexpected behaviour was observed in the CPU and memory intensive benchmarks, and in the networking benchmarks. The average overhead of running inside KVM is between zero and 30 percent when the host is lightly loaded (running only the system software and the necessary qemu-system-arm virtualization code), but the relative overhead increases when both host and VM is busy. We conjecture that this is related to the scheduling inside the host Linux.

A Knoppix-based demonstration environment for JPF

This paper explores how a KVM virtual machine booted with a Knoppix Live DVD can provide a simple and reliable system for sharing demonstrations of JPF, and in particular, for running regression tests in a repeatable way before changes are committed to shared repositories. To make the system easy to automate, we integrated host file system access and a server for script execution. To make it practical for an interactive workow, checkpointing was added to avoid booting and configuration delays. As an unexpected benefit, the isolation provided by the virtual machines allows multiple tests to run in parallel without risk of clashes over resources such as server ports.

Adaptive Online Compression in Clouds—Making Informed Decisions in Virtual Machine Environments

Infrastructure as a Service clouds often use virtual machines to host different customers on the same physical hardware. This form of resource sharing can lead to unpredicatable performance degradations for the individual customer, especially with regard to data-intensive applications, which heavily depend on stable I/O characteristics. One traditional approach to cope with I/O fluctuations is adaptive online compression. In this paper we present a new scheme for adaptive online compression which has been explicitly designed to work in co-located virtual machine environments. In contrast to existing adaptive online compression schemes, the decision model of our approach does not rely on the system metrics CPU utilization and I/O bandwidth, which we demonstrate to be often displayed inaccurately inside XEN, KVM, and Amazon EC2-based virtual machines. Instead, it only considers the application data rate. Without requiring any calibration or training phase our adaptive compression scheme can improve the I/O throughput of virtual machines up to a factor of four as shown through extended experimental evaluations.

Implementation of a Memory Disclosure Attack on Memory Deduplication of Virtual Machines

Memory deduplication improves the utilization of physical memory by sharing identical blocks of data. Although memory deduplication is most effective when many virtual machines with same operating systems run on a CPU, cross-user memory deduplication is a covert channel and causes serious memory disclosure attack. It reveals the existence of an application or file on another virtual machine. The covert channel is a difference in write access time on deduplicated memory pages that are re-created by Copy-On-Write, but it has some interferences caused by execution environments. This paper indicates that the attack includes implementation issues caused by memory alignment, self-reflection between page cache and heap, and run-time modification (swap-out, anonymous pages, ASLR, preloading mechanism, and self-modification code). However, these problems are avoidable with some techniques. In our experience on KSM (kernel samepage merging) with the KVM virtual machine, the attack could detect the security level of attacked operating systems, find vulnerable applications, and confirm the status of attacked applications.