Improving Network I/O Performance in Virtualized Clouds
Departmemt of Computer Science
Stuart Building 111
Wednesday, April 11th, 2012
Abstract: Virtualization is a key technology that powers cloud computing platforms such as Amazon EC2. Virtual machine (VM) consolidation, where multiple VMs share a physical host, has seen rapid adoption in practice with increasingly large number of VMs per machine and per CPU core. Our investigations, however, suggest that the increasing degree of VM consolidation has serious negative effects on the VMs TCP transport performance. As multiple VMs share a given CPU, the scheduling latencies, which can be in the order of tens of milliseconds, substantially increase the typically sub-millisecond round-trip times (RTTs) for TCP connections in a datacenter, causing significant degradation in transport throughput. To address this problem, I will present our integrated solution that consists of two complementary techniques: (1) vFlood that offloads a VM's TCP congestion control function to the driver domain for improving TCP transmit (TX) performance and (2) vSnoop that offloads the VM's TCP acknowledgement function to the driver domain for improving TCP receive (RX) performance. I will present evaluation results with our Xen-based prototypes, which demonstrate significant performance gain at both network transport and application levels.
Bio: Dongyan Xu is an associate professor in the Computer Science Department at Purdue University. His research focuses on the development of advanced virtualization technologies for cyber security and for cloud computing. He has also made early contribution to the area of peer-to-peer media streaming. He received an NSF CAREER Award in 2006 for his research towards virtual infrastructures on shared distributed platforms. His paper on optimizing virtual machine networking performance received the Paper of Distinction recognition at the 2011 ACM Symposium on Cloud Computing. His website is at http://www.cs.purdue.edu/homes/dxu/.