基于虚拟机的透明计算设备子系统设计及实现*

为解决在现有透明计算系统中磁盘等设备的虚拟化,需要对操作系统进行修改才能在端计算机上运行的问题,提出透明计算系统中一种基于虚拟机技术的设备子系统。以Intel VT硬件级虚拟支持和Xen虚拟化技术为基础, 在全虚拟化的虚拟机上运行用户操作系统: 通过运行在管理域用户空间中设备模型的虚拟磁盘和网络驱动,将用户域访问磁盘和网络的I/O请求跨网络地重定向到服务器进行处理, 从而实现端计算机上多操作系统的远程运行。在VT硬件平台和Xen虚拟机监控器上实现了原型系统,验证了该方法的可行性。     

基于同步机制的虚拟机磁盘在线迁移策略

提出一种带冗余检测的磁盘在线迁移策略。采用周期同步方式保持虚拟机在源、目的节点上磁盘的数据一致性,在非同步期内,利用位图标记源节点上被写的数据块(脏块)位置,在同步期内,检测脏块与目的节点相应块的Hash值,仅将不匹配的脏块同步至目的节 点,在磁盘数据一致后,启动虚拟机在线迁移。实验结果表明,该策略能减少磁盘迁移的数据量,降低磁盘迁移对高带宽的要求,虚拟机迁移时间与共享存储方式接近。     

基于优化的COW虚拟块设备的虚拟机按需部署机制

基于COW(Copy-on-Write)读写模式的虚拟块设备有利于实现大规模虚拟机环境下虚拟机的快速部署.文中为虚拟机管理器中的COW虚拟块设备设计了一种优化方法,能够提高COW磁盘的访问性能以及生成多个小尺寸的COW磁盘映像文件,以降低通过网络部署虚拟机的开销.基于优化的COW虚拟块设备,文中提出了虚拟机环境下的虚拟机按需部署机制及关键技术问题的解决方案.基于Linux平台和QEMU虚拟机,实现了基于优化COW虚拟块设备的虚拟机按需部署原型系统.实验表明,优化的COW虚拟块设备、基于COW磁盘有效工作集的优化部署以及COW磁盘回收等方法能够有效地支持虚拟机环境下低开销的、按需的虚拟机部署.     

适应广域网的虚拟机在线迁移模型

针对广域网(WAN)环境下虚拟机(VM)迁移存在的诸多问题提出一种迁移模型。该模型持续检测节点间链路状态,优化磁盘、内存和CPU状态及网络的迁移时序,并采用磁盘周期同步、单向网络隧道及虚拟机本地化等方法。实验表明：模型减少了虚拟机跨WAN迁移的数据量,缩短了网络重定向路径,在模拟实验中迁移的总时间及暂停时间接近于共享存储方式。     

面向虚拟机的远程磁盘缓存

在虚拟机(virtual machine)系统中,随着虚拟机数量和应用程序需求的不断增长,内存容量已经成为应用程序性能的主要瓶颈。为了提升内存密集型和I/O密集型程序的页面交换性能,提出了虚拟机的远程磁盘缓存机制REMOCA,它允许运行在一台物理主机上的虚拟机将其他物理主机的内存作为其二级磁盘缓存。由于网络传输延迟远远小于磁盘访问,用网络传输代替磁盘访问就能够有效地降低虚拟机的平均磁盘访问延迟。REMOCA的目标就要尽可能地减少磁盘访问。REMOCA运行在虚拟机管理器中,其基本工作原理是截获并处理虚拟机的页面淘汰、磁盘访问等事件。REMOCA能够与现有的虚拟机内存管理机制(如气球技术、影子缓存)相结合,从而提供更加灵活的内存资源管理策略。实验数据表明,REMOCA能有效地降低页面抖动对虚拟机性能的影响,并在很大程度上提升虚拟机中I/O密集型应用的性能。     

虚拟机环境下软件按需部署中的预取机制

针对大规模虚拟机环境下软件的按需部署,提出了一种基于预取的按需软件部署优化机制,能够降低用户端虚拟机的启动延迟以及为用户提供更好的虚拟机本地运行性能.基于用户使用软件的行为特点以及虚拟磁盘映像的细粒度分割,预取机制在后台对服务器端存储的虚拟磁盘映像进行预取,通过一种基于访问频率和优先级的预取目标识别算法AFPTR(access frequency and priority-based prefetch target recognition)和一种预取量动态调节机制,将预取集中在用户使用的少数小尺寸的虚拟磁盘映像上,并在预取过程中对预取量进行动态自适应地调节,以提高虚拟磁盘访问的本地命中率,进而提高用户端虚拟机的运行性能.基于QEMU虚拟机和Linux平台,实现了基于预取的按需软件部署原型系统.实验结果表明,预取机制能够有效地降低虚拟机的启动延迟,并能提高虚拟机的本地运行性能,支持虚拟机环境下按需、快速的软件部署.     

基于自动测试的虚拟机磁盘I/O服务质量评估框架

通过系统虚拟化技术在同一物理主机上的各个虚拟机之间动态地分配物理资源,可以实现物理资源的最大利用率和分配的公平性,达到较高的服务质量.提出了一种基于测试和汇总同一物理主机上各虚拟机之间物理磁盘I/O带宽的分配情况,来评估虚拟机磁盘I/O服务质量的框架.它通过客户端控制各虚拟机执行指定的I/O测试程序,同步收集各虚拟机的虚拟磁盘吞吐量,并加以汇总.实验证明这套框架可以有效地从高效性和公平性两方面评估虚拟化环境中虚拟机磁盘I/O的服务质量.     

一种高效的虚拟机快照管理系统

链式虚拟磁盘快照技术是广泛实现的一种支持虚拟机失效恢复的技术手段,针对链式结构磁盘快照技术引入多余I／O请求导致虚拟机性能低效问题,分析和研究了支持高效虚拟机快照实现的无链式关联磁盘快照技术,通过集成ZFS及OCFS2文件系统给出了无链式关联磁盘快照技术实现,并设计和实现了虚拟机快照存储组织模型和快速检索算法,提高了虚拟机快照检索效率．设计试验验证了系统的有效性．     

一种高效的虚拟机磁盘快照系统

提出了一种带缓存的虚拟机磁盘快照系统(LiveDiskSnap),以支持不停机的虚拟机磁盘快照.通过在虚拟机监控器层加入了缓存块设备驱动,并采用写时重定向技术,使得快照中无需中断虚拟机的运行,同时保证了快照前后虚拟机磁盘状态的一致性.实验结果表明,LiveDiskSnap可在虚拟机不停机的情况下进行磁盘快照,且快照过程中的平均性能损失小于8％.     

基于KVM的虚拟锁步技术

依托基于内核的虚拟机（kernel-based virtual machine , KVM ）的平台,针对虚拟机容错系统中的关键技术---虚拟锁步技术展开研究,主要研究虚拟锁步技术所采用的虚拟机同步机制（VM synchronization mechanism ）。对开源虚拟机容错软件Kemari进行架构剖析与代码分析,指出其所使用的基于数据拷贝的虚拟机同步机制在不使用共享存储进行锁步运行时,具有一定性能缺陷;以此为基础,提出相应的改进措施,设计并实现一种新的虚拟机同步机制。该机制采用事件重放的方式实现冗余虚拟机间块设备数据的同步,弥补了Kemari虚拟机同步机制的相关性能缺陷。     

Chunk mode VM migration in XIA and triple-way pipeline for performance optimization

Traditional TCP/IP network is showing lots of shortages and research for future networks is becoming a hotspot. FIA (Future Internet Architecture) and FIA-NP (Next Phase) are supported by US NSF for future Internet designing. Moreover, virtual machine migration is a significant technique in cloud computing. As a network application, it should also be supported in XIA (eXpressive Internet Architecture), which is in both FIA and FIA-NP projects. Current research just achieved to conduct VM migration in XIA, but the performance is not satisfactory. In this paper, we firstly present three methods including DAG design and routing table modification to maintain VM’s connectivity and communication states, concerning performance and overhead. VM migration experiments are conducted intra-AD, inter-AD and across IP network with KVM instances. The procedure is achieved by a migration control protocol which is suitable for the characters of XIA. Secondly, we propose a triple-way pipeline to improve the performance of chunk mode data transfer and greatly reduce total migration time and downtime. Evaluation results show that our solutions can well supports full live VM migration over XIA network respectively, keeping services seamless and performance improved. Furthermore, the optimization scheme can greatly improve the speed of VM migration, especially on XIA testbed.     

虚拟机实时迁移技术研究

分析了虚拟机的实时迁移原理,重点论述了Xen的实时迁移实现方法,并在不同负载情况下对Xen的实时迁移进行了实验和探讨。     

Resource allocation algorithms for virtualized service hosting platforms

Commodity clusters are used routinely for deploying service hosting platforms. Due to hardware and operation costs, clusters need to be shared among multiple services. Crucial for enabling such shared hosting platforms is virtual machine (VM) technology, which allows consolidation of hardware resources. A key challenge, however, is to make appropriate decisions when allocating hardware resources to service instances. In this work we propose a formulation of the resource allocation problem in shared hosting platforms for static workloads with servers that provide multiple types of resources. Our formulation supports a mix of best-effort and QoS scenarios, and, via a precisely defined objective function, promotes performance, fairness, and cluster utilization. Further, this formulation makes it possible to compute a bound on the optimal resource allocation. We propose several classes of resource allocation algorithms, which we evaluate in simulation. We are able to identify an algorithm that achieves average performance close to the optimal across many experimental scenarios. Furthermore, this algorithm runs in only a few seconds for large platforms and thus is usable in practice.     

High performance network virtualization with SR-IOV

Virtualization poses new challenges to I/O performance. The single-root I/O virtualization (SR-IOV) standard allows an I/O device to be shared by multiple Virtual Machines (VMs), without losing performance. We propose a generic virtualization architecture for SR-IOV-capable devices, which can be implemented on multiple Virtual Machine Monitors (VMMs). With the support of our architecture, the SR-IOV-capable device driver is highly portable and agnostic of the underlying VMM. Because the Virtual Function (VF) driver with SR-IOV architecture sticks to hardware and poses a challenge to VM migration, we also propose a dynamic network interface switching (DNIS) scheme to address the migration challenge. Based on our first implementation of the network device driver, we deployed several optimizations to reduce virtualization overhead. Then, we conducted comprehensive experiments to evaluate SR-IOV performance. The results show that SR-IOV can achieve a line rate throughput (9.48 Gbps) and scale network up to 60 VMs, at the cost of only 1.76% additional CPU overhead per VM, without sacrificing throughput and migration.     

An Efficient On-Demand Virtual Machine Migration in Cloud Using Common Deployment Model

Cloud Computing provides various services to the customer in a flexible and reliable manner. Virtual Machines (VM) are created from physical resources of the data center for handling huge number of requests as a task. These tasks are executed in the VM at the data center which needs excess hosts for satisfying the customer request. The VM migration solves this problem by migrating the VM from one host to another host and makes the resources available at any time. This process is carried out based on various algorithms which follow a predefined capacity of source VM leads to the capacity issue at the destination VM. The proposed VM migration technique performs the migration process based on the request of the requesting host machine. This technique can perform in three ways namely single VM migration, Multiple VM migration and Cluster VM migration. Common Deployment Manager (CDM) is used to support through negotiation that happens across the source host and destination host for providing the high quality service to their customer. The VM migration requests are handled with an exposure of the source host capabilities. The proposed analysis also uses the retired instructions with execution by the hypervisor to achieve high reliability. The objective of the proposed technique is to perform a VM migration process based on the prior knowledge of the resource availability in the target VM.     

虚拟计算环境下虚拟机资源负载均衡方法

针对虚拟机资源粒度大和迁移时传输数据量大的特点,提出一种基于虚拟机迁移的负载均衡方法。该方法利用负载阈值对宿主机后续时间节点的负载趋势进行预测,避免瞬时负载峰值触发的虚拟机迁移问题。在触发迁移后采用加权概率转发方式选择迁移目标节点,解决传统负载均衡技术中的群聚冲突问题。实验结果表明,在宿主机负载分布严重不平衡的情况下,该方法能有效改善系统性能。     

Multi-objective dynamic virtual machine consolidation in the cloud using ant colony system

In this paper, we present a novel multi-objective ant colony system algorithm for virtual machine (VM) consolidation in cloud data centres. The proposed algorithm builds VM migration plans, which are then used to minimise over-provisioning of physical machines (PMs) by consolidating VMs on under-utilised PMs. It optimises two objectives that are ordered by their importance. The first and foremost objective in the proposed algorithm is to maximise the number of released PMs. Moreover, since VM migration is a resource-intensive operation, it also tries to minimise the number of VM migrations. The proposed algorithm is empirically evaluated in a series of experiments. The experimental results show that the proposed algorithm provides an efficient solution for VM consolidation in cloud data centres. Moreover, it outperforms two existing ant colony optimization-based VM consolidation algorithms in terms of number of released PMs and number of VM migrations.     

DataABC: A fast ABC based energy-efficient live VM consolidation policy with data-intensive energy evaluation model

Live Virtual Machine (VM) consolidation is an effective method of improving energy-efficiency level in green data centers. Currently, to evaluate energy consumption in green data centers, energy-efficiency evaluation model with CPU utilization rate has been proposed. However, it is not suitable for data-intensive computing due to great energy consumption by GPU-intensive processing. In this paper, we have proposed a new energy evaluation model with CPU and GPU utilization rates. There are two kinds of policies in live VM consolidation: one for VM selection and the other for VM allocation. Some researchers have proposed their solutions based on VM selection policy or VM allocation policy respectively. However, it will be a better energy-efficiency VM consolidation policy if these two polices are integrated together. Based on these two policies, a fast Artificial Bee Colony (ABC) based energy-efficiency live VM consolidation policy with data-intensive energy model, named as DataABC, is proposed. DataABC adopts the idea of Artificial Bee Colony algorithm to get a fast and global optimized decision of VM consolidation. Compared with two state-of-art policies of PS-ABC and PS-ES, the total energy consumption of DataABC evidently drop by 9.72% and 5.84% respectively. As a result, based on the ESV metric, the DataABC approach has proved that (a) the energy-efficiency evaluation model with data-intensive computing is valid and that (b) DataABC can save energy with a good Quality of Service (QoS) in green data centers.