共生关系感知的虚拟机间通信优化技术研究

VMM technology by design enables VMs (Virtual Machines) co-located on the same physical host to share resources and at the same time treats all VMs as independent computing nodes regardless of where these VMs are located. VMM also supports VM live migration. All these features make VM widely deployed into datacenters and cloud computing environments. However, network I/O is known to be one of the dominating workloads in virtualized clouds. One way to improve inter-VM communication efficiency is to support co-located VM communication by using a faster shared-memory based communication channel than the traditional TCP/IP commonly used for inter-VM communications regardless whether VMs are co-located or not. Although existing co-located VM communication mechanisms are proposed to reduce the unnecessary long path through the TCP/IP network stack, however, few of them provides adaptive transparent VM live migration support. And shared memory is not managed in an efficient and flexible way. The utility of shared memory is not optimized. In this project, we will focus on how to benefit from the co-location of VMs by introducing residency-aware inter-VM communication technology. Aiming at issues addressed above, first, we will establish multi-goal driven inter-VM communication optimizing model and explore on how to reduce the communication overhead. Then, we will research on VM live migration detection, dynamic co-located VM membership maintaining algorithm, and adaptive transparent switching method for local shared memory channel and traditional TCP/IP channel to provide VM live migration and dynamic deployment support. Third, we will study on how to provide workload adaptive and elastic shared memory allocation and revocation. Last but not least, we will validate the efficiency and correctness of proposed methods and algorithms with experiments. The research work in this project is important not only for its theoretical value but also for its promising prospect of being widely used.

虚拟机间通信效率已成为虚拟云平台中网络密集型应用的主要性能瓶颈之一。一种普遍而有效的优化方法是采用共享内存支持共生虚拟机间通信。现有方法对虚拟机在线迁移支持不足，存在共享内存利用率低及无法动态调节的问题。针对上述问题，本项目将研究高效、自适应的共生关系感知的虚拟机间通信优化技术，重点突破自适应的虚拟机在线迁移支持、负载感知的共享内存按需动态分配和管理等技术，以进一步提高通信效率、支持虚拟机在线迁移和动态部署、提高共享内存利用率。将结合虚拟云平台应用、通过实验验证上述方法的正确性和有效性。项目研究具有重要的理论价值和广泛的应用前景。

虚拟机间通信效率已成为虚拟云平台中网络密集型应用的主要性能瓶颈之一。一种普遍而有效的优化方法是采用共享内存支持共生虚拟机间通信。现有方法不支持基于多核CPU的通信优化，对虚拟机在线迁移支持不足，共享内存利用率低且无法动态调节，安全可信有待增强。针对上述问题，本项目重点研究了多目标驱动的共生关系感知的虚拟机间通信优化模型、自适应的虚拟机在线迁移与动态部署支持、多虚拟机间共享内存的弹性管理及安全可信增强等技术，在改善通信性能、支持虚拟机在线迁移、提高共享内存利用率、增强安全可信方面取得突破。本项目主要贡献包括：.第一，面向通信效率优化、虚拟机在线迁移和动态部署支持、多层透明性三个目标，研究了共享内存通道所在软件栈的层次与通信效率及透明性间的关联，确定了软件栈层次和设计策略，建立了多目标驱动的共生关系感知的虚拟机间通信优化模型及原型系统，并对其网络I/O性能、虚拟机在线迁移支持及透明性进行了验证；.第二，针对虚拟机在线迁移支持这一难点技术，深入研究了自适应的虚拟机在线迁移与动态部署支持技术，突破了虚拟机在线迁移感知技术，提出了共生虚拟机的动态发现、面向双通信模式的自适应切换等方法，验证了方法的正确性和有效性；.第三，针对虚拟机间共享内存利用率低、安全可信能力有待提高等问题，提出了负载感知的共享内存按需动态分配和管理方法，在多虚拟机的安全可信增强技术上取得突破，验证了所提方法及关键机制的有效性。.项目按照研究计划展开研究，达到了预期研究目标，形成了预期研究成果。项目研究过程中，已录用/发表论文共8篇，包括CCF A类国际会议1篇、CCF C类国际会议2篇，软件学报等国内外期刊论文5篇，已有3篇被SCI检索、2篇被EI检索，授权专利3项，构建了共生关系感知的虚拟机间通信优化原型系统，验证了本项目所提出方法的高效、自适应、安全等特性。项目研究具有重要的理论价值和广泛的应用前景。