Huadong Dai

Residency-Aware Virtual Machine Communication Optimization: Design Choices and Techniques

Network I/O workloads are dominating in many data centers and cloud computing environments today. One way to improve inter Virtual Machine (VM) communication efficiency is to support co-resident VM communication by using shared memory based approaches and to resort to the traditional TCP/IP for inter-VM communications between VMs that are located on different physical hosts. Although a number of independent efforts are dedicated to improving communication efficiency between co-resident VMs, they differ from one another in terms of how the inter-VM communication optimization is carried out and where in the software stack the shared memory channel is established. In this paper, we provide an in-depth overview of the design choices and techniques for optimizing the performance of the co-resident inter-VM communication, with dual objectives. First, we describe the core design guidelines and key issues for optimizing inter-VM communication by using shared memory based mechanisms. Typical issues include choices of implementation layer in the software stack, seamless agility for VM live migration and VM dynamic deployment support, multilevel transparency. Second, we conduct a comprehensive analysis of representative state-of-the-art research efforts and implementation techniques based on the core design guidelines. We also give an analysis of future requirements in advanced features such as reliability, security and stability. The research reported in this paper not only provides the reference for developing the next generation of inter-VM communication optimization mechanisms, but also offers opportunities for both cloud infrastructure providers and cloud service consumers to improve inter-VM communication efficiency in virtualized platforms.

Shared-Memory Optimizations for Inter-Virtual-Machine Communication

Virtual machines (VMs) and virtualization are one of the core computing technologies today. Inter-VM communication is not only prevalent but also one of the leading costs for data-intensive systems and applications in most data centers and cloud computing environments. One way to improve inter-VM communication efficiency is to support coresident VM communication using shared-memory-based methods and resort to the traditional TCP/IP for communications between VMs that are located on different physical machines. In recent years, several independent kernel development efforts have been dedicated to improving communication efficiency between coresident VMs using shared-memory channels, and the development efforts differ from one another in terms of where and how the shared-memory channel is established. In this article, we provide a comprehensive overview of the design choices and techniques for performance optimization of coresident inter-VM communication. We examine the key issues for improving inter-VM communication using shared-memory-based mechanisms, such as implementation choices in the software stack, seamless agility for dynamic addition or removal of coresident VMs, and multilevel transparency, as well as advanced requirements in reliability, security, and stability. An in-depth comparison of state-of-the-art research efforts, implementation techniques, evaluation methods, and performance is conducted. We conjecture that this comprehensive survey will not only provide the foundation for developing the next generation of inter-VM communication optimization mechanisms but also offers opportunities to both cloud infrastructure providers and cloud service providers and consumers for improving communication efficiency between coresident VMs in virtualized computing platforms.

SPOTOS: A Safe Point Detector for Online Patching Commodity OS

For high availability systems, it is desirable to fix discovered operating system security holes and other bugs, and do performance improvements without rebooting. However, the way that commodity operating systems are implemented sometimes makes online patching rather difficult, especially when the code to be updated is frequently used by the kernel. For commodity OS, hot update may have to be aborted due to not finding a safe point, even if there do exist occasions that those codes are not in use. After analyzing the difficulties faced when detecting a commodity OS’ quiescent state, this article proposes using reference counting copy to dynamically monitor the code execution status, which can trigger the online patching whenever safe point is reached. In principle, the proposed approach is applicable to all the legacy operating systems that support dynamically loading kernel modules, while minor changes will improve the performance and decrease implementation complexity. Based on Ksplice, a working prototype, SPOTOS, is implemented on Linux 2.6. SPOTOS has the advantage of dynamically monitoring the execution status of all the functions that are to be updated. Experiments on network communication are conducted, and the results prove the effectiveness and efficiency of our approach.

Enhanced Non-Blocking Buffer: A High Performance Non-Blocking Message Communication Mechanism

In high performance computing, it is desirable for threads to communicate with each other without incurring nonessential mutual exclusion. Various non-blocking mechanisms are proposed in the literature. However, the previous studies have the problem of message overwritten or repeatedly read. And none of them can handle the dispatcher/workers processing model well. Utilizing the primitive Compare and Swap, which can be executed atomically on most hardware platforms, the Enhanced Nonblocking Buffer (ENBB) mechanism described in this article is free of these drawbacks. The algorithms and their proofs are also presented. Lastly, experiments show the ENBB mechanism is both bug-free and high performance. Keywordsconcurrent programming, mutual exclusion, blocking, producer, consumer