Building an Effective and Flexible Event Path for I/O Virtualization with ES2 PROJECT TITLE : ES2: Building an Efficient and Responsive Event Path for I/O Virtualization ABSTRACT: I/O virtualization's primary performance bottleneck is hypervisor intervention in the virtual I/O event path. This is due to the fact that hypervisor intervention results in costly VM exits. The inadequacies of earlier software solutions against the delivery of virtual interrupts, a major cause of VM exits, led to the development of the hardware-based Posted-Interrupt (PI) technology. PI has the capability of delivering non-exit interrupts without jeopardizing any of the benefits of virtualization. However, it only affects one half of the event path, which is the interrupt path; guests' I/O requests can still cause a significant number of VM exits. This is because it only acts on the interrupt path. In addition, while delivering interrupts, PI may still experience a significant amount of latency as a result of the vCPU scheduling. We propose ES2 in the hopes of achieving an ideal event flow in order to simultaneously enhance the delivery of bidirectional I/O events between guests and the devices they use. The PI serves as the foundation for the hybrid I/O handling scheme that ES2 introduces. This scheme allows for efficient I/O request delivery and intelligent interrupt redirection, which together improve I/O responsiveness. It does not call for any modifications to be made to the guest operating system. We show that ES2 significantly cuts down on I/O-related VM exits by lowering the exit handling time (EHT) to below 2.5 percent for TCP streams and below 0.1 percent for UDP streams. Additionally, we show that ES2 significantly boosts guest throughput by 1.9 times for Memcached and by 1.6 times for Nginx while maintaining a low level of guest latency. Did you like this research project? To get this research project Guidelines, Training and Code... Click Here facebook twitter google+ linkedin stumble pinterest Exploring Binary Code Sharing's Potential and Possibility in Mobile Computing scheduling energy-conscious cloud workflow applications using geographically distributed data