Scheduling I/O in virtual machine monitors

Abstract

This paper explores the relationship between domain scheduling in a virtual machine monitor (VMM) and I/O performance. Traditionally, VMM schedulers have focused on fairly sharing the processor resources among domains while leaving the scheduling of I/O resources as a secondary concern. However, this can result in poor and/or unpredictable application performance, making virtualization less desirable for applications that require efficient and consistent I/O behavior.This paper is the first to study the impact of the VMM scheduler on performance using multiple guest domains concurrently running different types of applications. In particular, different combinations of processor-intensive, bandwidth-intensive, and latencysensitive applications are run concurrently to quantify the impacts of different scheduler configurations on processor and I/O performance. These applications are evaluated on 11 different scheduler configurations within the Xen VMM. These configurations include a variety of scheduler extensions aimed at improving I/O performance. This cross product of scheduler configurations and application types offers insight into the key problems in VMM scheduling for I/O and motivates future innovation in this area.

Coexisting Scheduling Policies Boosting I/O Virtual Machines

“…Ongaro et al [7] examine the Xen's Credit Scheduler and expose its vulnerabilities from an I/O performance perspective. The authors evaluate two basic existing features of Credit and propose run-queue sorting according to the credits each VM has consumed.…”

“…The authors evaluate two basic existing features of Credit and propose run-queue sorting according to the credits each VM has consumed. Contrary to our approach, based on multiple, co-existing scheduling policies, the authors in [7] optimize an existing, unified scheduler to favor I/O VMs.…”

“…Credit's Shortcomings: As a general purpose scheduler, Credit as expected falls shorts in some cases. If a VM yields the processor before accounting occurs, no credits are debited [7]. This gives the running VM an advantage over others that run for a bit longer.…”

Providing Performance Guarantees to Virtual Machines Using Real-Time Scheduling

“…In [11], the authors proposed to augment the Xen hypervisor with a set of mechanisms to account for and to control the CPU time spent on behalf of VMs doing I/O. In [17], the authors proposed an extension to the Xen credit-based scheduler improving its behavior in presence of multiple different applications with heavy I/O workloads, prioritizing the I/O bound ones. Also, in [13], the authors proposed to modify the Xen CPU scheduler and networking architecture to improve the performance of virtualized I/O on 10 Gbps Ethernet.…”

Analyzing and Improving MPI Communication Performance in Overcommitted Virtualized Systems

“…The biggest shortcoming of the Credit scheduler is its poor support for I/O events. The improvement works, such as the introductions of the BOOST priority [17] and BCredit [18], which essentially works by changing the dom0 VCPUs' credits so that dom0 remains in UNDER state and keeps its BOOST priority, have been made to alleviate the problem.…”