Analyzing and Improving MPI Communication Performance in Overcommitted Virtualized Systems

Shao, Zili; Wang, Qiang; Xie, Xuejiao; Jin, Hai; He, Ligang

doi:10.1109/mascots.2011.27

Cited by 8 publications

(2 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We choose three benchmark programs (called is, ep, and lu) from NPB suite of version 2.4 [15], as they exhibit three typical types of parallel executions: communication intensive application with little computation (is); CPU intensive application with little communication (ep); and the one that lies in between them (lu) [16]. For each VM, the number of packets and that of spinlocks are recorded every 120 milliseconds (multiplying the 30ms of Xen Credit scheduler by the number of VMs in this experiment) over 60 seconds.…”

Section: Basic Ideamentioning

confidence: 99%

“…Four functions are used to get a suitable VCPU when runnable VCPUs include parallel ones (lines [8][9][10][11][12][13][14][15][16][17][18][19]. First, it calls get_parallel_vcpus_with_maxPackets() to obtain VCPUs, which are put in vcpuSet1, from run queue runq if parallel VMs they belongs to receive maximum packets since their last de-scheduled (line 8).…”

Section: Svs Algorithmmentioning

confidence: 99%

See 1 more Smart Citation

Synchronization-Aware Scheduling for Virtual Clusters in Cloud

Chen

et al. 2015

IEEE Trans. Parallel Distrib. Syst.

Self Cite

View full text Add to dashboard Cite

Due to high flexibility and cost-effectiveness, cloud computing is increasingly being explored as an alternative to local clusters by academic and commercial users. Recent research already confirmed the feasibility of running tightly-coupled parallel applications with virtual clusters. However, such types of applications suffer from significant performance degradation, especially as the overcommitment is common in cloud. That is, the number of executable Virtual CPUs (VCPUs) is often larger than that of available Physical CPUs (PCPUs) in the system. The performance degradation is mainly due to the fact that the current Virtual Machine Monitors (VMMs) are unaware of the synchronization requirements of the VMs which are running parallel applications. In this paper, There are two key contributions. (1) We propose an autonomous synchronization-aware VM scheduling (SVS) algorithm, which can effectively mitigate the performance degradation of tightly-coupled parallel applications running atop them in overcommitted situation.(2) We integrate the SVS algorithm into Xen VMM scheduler, and rigorously implement a prototype. We evaluate our design on a real cluster environment with NPB benchmark and real-world trace. Experiments show that our solution attains better performance for tightly-coupled parallel applications than the state-of-the-art approaches like Xen's Credit scheduler, balance scheduling, and hybrid scheduling.

show abstract

Section: Basic Ideamentioning

confidence: 99%

Section: Svs Algorithmmentioning

confidence: 99%

Synchronization-Aware Scheduling for Virtual Clusters in Cloud

Chen

et al. 2015

IEEE Trans. Parallel Distrib. Syst.

Self Cite

View full text Add to dashboard Cite

show abstract

CCHybrid: CPU co‐scheduling in virtualization environment

2017

Concurrency and Computation

View full text Add to dashboard Cite

SummaryVirtualization is very important to build the emerging cloud infrastructure, and a VM (virtual machine) with many kinds of workloads can run on physical machines in cloud environment. The VMM (virtual machine manager) scheduling algorithm asynchronously schedules each VCPU (virtual CPU) of a VM and ensures the CPU time usage of each VM. This proportional share method is widely used, because it simplifies the implementation of VMM CPU scheduling algorithm and can provide near‐perfect performance for most ordinary workloads. However, when a VM runs with parallel workloads, the above method causes performance degradation because of the negative impact of virtualized systems. Therefore, in this paper, we propose an optimized scheduling system, called CCHybrid, for parallel program in the Xen. It uses weight‐based proportion share strategy to ensure the fairness. In order to resolve the impact of virtualization on synchronization, it uses a novel co‐scheduling strategy, which dynamically adjusts the size of co‐scheduling to remit CPU fragmentation and maintains the original asynchronous scheduling policy for non‐parallel applications. In this way, CCHybrid provides CPU resource allocation services for Xen and can decrease the negative impact of virtualized systems, while ensuring the fairness of VMs and the performance of non‐parallel workload. Experimental results show that in the case of multiple VMs, CCHybrid improves the performance of parallel workload from 15% to 50%, and the impact on non‐parallel workload is less than 5%, in comparison with the credit scheduling algorithm of Xen.

show abstract

vPlacer: A Co-scheduler for Optimizing the Performance of Parallel Jobs in Xen

Jiang

Ren

et al. 2018

Algorithms and Architectures for Parallel Processing

View full text Add to dashboard Cite

Analyzing and Improving MPI Communication Performance in Overcommitted Virtualized Systems

Cited by 8 publications

References 15 publications

Synchronization-Aware Scheduling for Virtual Clusters in Cloud

Synchronization-Aware Scheduling for Virtual Clusters in Cloud

CCHybrid: CPU co‐scheduling in virtualization environment

vPlacer: A Co-scheduler for Optimizing the Performance of Parallel Jobs in Xen

Contact Info

Product

Resources

About