Revisiting Performance Interference among Consolidated n-Tier Applications: Sharing is Better Than Isolation

Abstract

Abstract-Performance unpredictability is one of the major concerns slowing down the migration of mission-critical applications into cloud computing infrastructures [4]. An example of non-intuitive result is the measured n-tier application performance in a virtualized environment that showed increasing workload caused a competing, co-located constant workload to decrease its response time [12]. In this paper, we investigate the sensitivity of measured performance in relation to two factors: (1) consolidated server specification of virtual machine resource availability, and (2) burstiness of n-tier application workload. Our first and surprising finding is that specifying a complete isolation, e.g., 50-50 even split of CPU between two co-located virtual machines (VMs) results in significantly lower performance compared to a fully-shared allocation, e.g., up to 100% CPU for both co-located VMs. This happens even at relatively modest resource utilization levels (e.g., 40% CPU in the VMs). Second, we found that an increasingly bursty workload also increases the performance loss among the consolidated servers, even at similarly modest utilization levels (e.g., 70% overall). A potential solution to the first problem (performance loss due to resource allocation) is crosstier-priority scheduling (giving higher priority to shorter jobs), which can reduce the performance loss by a factor of two in our experiments. In contrast, bursty workloads are a more difficult problem: our measurements show they affect both the isolation and sharing strategies in virtual machine resource allocation.

Scheduling the I/O of HPC Applications Under Congestion

“…In distributed computing, the problem of having I/O NETWORK • I/O re-routing [19], [20], [21], [22] • Application-side I/O scheduling [1], [ performance variability due to sharing resources is wellknown and studied. Numerous papers analyze this problem for clouds [23], [24], [25]. In [24], Pu et al present a study of interference specifically for I/O workloads in the cloud in order to understand the performance factors that impact the efficiency and effectiveness of resource multiplexing and scheduling among VMs.…”

“…In [24], Pu et al present a study of interference specifically for I/O workloads in the cloud in order to understand the performance factors that impact the efficiency and effectiveness of resource multiplexing and scheduling among VMs. In [25], the authors investigate the sensitivity of measured performance in relation to consolidated server specification of virtual machine resource availability, and burstiness of n-tier application workload. Their results show that an increasingly bursty workload also increases the performance loss among the consolidated servers, however, without being able to offer a solution.…”

A Hardware/Software Approach for Mitigating Performance Interference Effects in Virtualized Environments Using SR-IOV

“…In a contrary approach, Kanemasa et al [14] demonstrated that shortcomings of strict isolation can outweigh its benefits. They show that a 50-50 split of CPU resources between two concurrent VMs can yield lower performance than a fully-shared allocation (100% CPU for both).…”

“…Some studies also address multiple resources [9]- [11]. Proposed techniques for mitigating performance interference effects cover isolation of shared resources between concurrent VMs [1]- [3], [12], [13] or sharing-based resource allocations [5], [14].…”

Vanguard

“…A datacenter customer running the same VM on the same physical host at different times is likely to experience wide disparity in performance, depending on the resource consumption by other VMs that happen to be co-located on this host [11,15]. Conservatively, and at significant cost, datacenter operators elect to keep physical servers at low utilization levels, to minimize the possibility of adverse performance interactions between customer workloads.…”