Performance Interference of Virtual Machines: A Survey

Abstract

The rapid development of cloud computing with virtualization technology has benefited both the academia and the industry. For any cloud data center at scale, one of the primary challenges is how to effectively orchestrate a large number of virtual machines (VMs) in a performance-aware and cost-effective manner. A key problem here is that the performance interference between VMs can significantly undermine the efficiency of cloud data centers, leading to performance degradation and additional operation cost. To address this issue, extensive studies have been conducted to investigate the problem from different aspects. In this survey, we make a comprehensive investigation into the causes of VM interference and provide an in-depth review of existing research and solutions in the literature. We first categorize existing studies on interference models according to their modeling objectives, metrics used and modeling methods. Then we revisit interference-aware strategies for scheduling optimization as well as co-optimization based approaches. Finally, the survey identifies open challenges with respect to VM interference in data centers and discusses possible research directions to provide insights for future research in the area.

gPerfIsol: GNN-Based Rate-Limits Allocation for Performance Isolation in Multi-Tenant Cloud

“…However, this only handles one aspect of isolation, namely, the logical isolation of traffic when it comes to security concerns [18]. It does not guarantee nor prevent performance interference between tenants [19] [20] [21]. Solving performance isolation requires creating new network-level isolation techniques with a tenant-centric perspective.…”

Systematic Cloud-Based Optimization: Twin-Fold Moth Flame Algorithm for VM Deployment and Load-Balancing

“…The database that was studied allows for the identification of several kinds of blocks, along with the attributes that are associated with each category: Category 1: Comprises blocks of dimensions [4,4,6,7], comprising 20 blocks, denoted as [T4, T1, T3, T2].…”

“…The current study seeks to improve the optimization of cloud-based computing by building upon extensive research. The most recent study has focused on using complex optimization methods to enhance virtual machine (VM) placement and load distribution, with significant contributions from [3,4]. Expanding on this path, the work introduces a new method and evaluates its effectiveness using indicators such as energy efficiency, lifespan analysis, and resource expenditures.…”

Performance interference of co-allocated applications: a systematic literature review

“…There are only a few efforts that have carried out a survey of performance interference approaches. Specifically, the works in [21] and [22] perform an analysis of the methods used to predict/analyze performance interference. The main difference of our work with these articles is that these works mainly focus on virtualized environments while we consider other platforms such as bare metal executions.…”