XCollOpts: A Novel Improvement of Network Virtualizations in Xen for I/O-Latency Sensitive Applications on Multicores

Abstract

It has long been recognized that the Credit scheduler selectively favors CPU-bound applications whereas for I/O-latency sensitive workloads, such as those related to stream-based audio/video services, it only exhibits tolerable, or even worse, unacceptable performance. The reasons behind this phenomenon are the poor understanding (to some degree) of the virtual machine scheduling as well as the network I/O virtualizations. In order to address these problems and make the system more responsive to the I/O-latency sensitive applications, in this paper, we present XCollOpts which performs a collection of novel optimizations to improve the Credit scheduler and the underlying I/O virtualizations in multicore environments, each from two perspectives. To optimize the schedule, in XCollOpts, we first pinpoint the Imbalanced Multi-Boosting problem among the cores thereby minimizing the system response time by load balancing the BOOST VCPUs. Then, we describe the Premature Preemption problem and address it by monitoring the received network packets in the driver domain and deliberately preventing it from being prematurely preempted during the packet delivery. However, these optimizations on the scheduling strategies cannot be fully exploited if the performance issues of the underlying supportive communication mechanisms are not considered. To this end, we make two further optimizations for the network I/O virtualizations, namely, Multi-Tasklet Pairs and Optimized Small Data Packet. Our empirical studies show that with XCollOpts, we can significantly improve the performance of the latency-sensitive applications at a cost of relatively small system overhead.Index Terms-Scheduling, hypervisor, multicore, load balancing, I/O virtualization.

Interrupt Latency Accurate Measurement in Multiprocessing Embedded Systems by Means of a Dedicated Circuit

“…It marks these I/O requests by the event channels using a technique called hypercall. When the domain is scheduled, it checks the event channels and delivers the pending events by calling the corresponding interrupt handler [9].…”

“…Each core can create eight vCPUs. Xen's Credit Scheduler synchronizes all the vCPUs with a fair share algorithm based on proportional scheduling [9]. Xen usually allocates just one vCPU to each domain, thus containing the information related to scheduling and event channels.…”

Fine-Grained I/O Traffic Control Middleware for I/O Fairness in Virtualized System

“…If the vCPU occupying the pCPU has OVER or UNDER priority, it preempts the pCPU for 10 ms. The BOOST mechanism improves I/O responsiveness to enhance the I/O performance of data-intensive VMs; however, the same BOOST priority between vCPUs causes an I/O performance imbalance [10], [31]. In addition, the round-robin scheduling policy defines the limits on the I/O performance improvement of the Credit1 scheduler [13], [32].…”

Guest Editors' Introduction: Special issue on efficient management of SDN/NFV-based systems - Part II

“…In "XCollOpts: A Novel Improvement of Network Virtualization in Xen for I/O-Latency Sensitive Applications on Multicores," Zeng et al [9] present several improvements to the Xen scheduler to make it more suitable for I/O-latency sensitive applications in virtualized multi-core environments, such as multimedia streaming applications. The imbalanced multi-boosting problem among the cores and the premature preemption problem are addressed, together with two further optimizations for the network I/O virtualizations.…”