Adjustable Credit Scheduling For High Performance Network Virtualization

Virtualization technology is now widely adopted in cloud computing to supportheterogeneous and dynamic workload. High-speed network interface is also widelyused in various cloud platforms, Platform network performance between virtual ma-chinesoftenhaveahugeimpactontheoverallperformanceofthecloudplatform. there-fore,improvingtheperformanceofhigh-speednetworkI/Ovirtualizationhasgreatsig-nifcance in promoting the development of coud computing. The scheduler in a virtualmachine monitor (VMM) plays an important role in allocating resources. However,the type of applications in virtual machines (VM) is unknown to the scheduler, andI/O-intensive and CPU-intensive applications are treated the same. This makes virtualsystemsunabletotakefulladvantageofhighperformancenetworkssuchas10-GigabitEthernet.In this paper, we review the SR-IOV networking solution and show by experimentthat the current credit scheduler in Xen does not utilize high performance networksefciently. We identify the two main reasons, the frst one is that the I/O-intensivevirtual machine can not get enough credit to handle I/O when the system’s load ishigh; the second one is that I/O-intensive domains are not scheduled timely. For thesereasons, we propose a novel scheduling model with two optimizations to eliminate thebottleneck caused by scheduler. In this model, guest domains are divided into I/O-intensive domains and CPU-intensive domains according to their monitored behavior.I/O-intensive domains can obtain extra credits that CPU-intensive domains are willingto share. Besides, the total number of credits available is adjusted agilely to acceleratethe I/O responsiveness. Our experimental evaluation with benchmarks shows that thenew scheduling model improves bandwidth even when the system load is very high as well as the fairness of scheduling.