Distributed resource management across process boundaries

Abstract

Multi-tenant distributed systems composed of small services, such as Service-oriented Architectures (SOAs) and Micro-services, raise new challenges in attaining high performance and efficient resource utilization. In these systems, a request execution spans tens to thousands of processes, and the execution paths and resource demands on different services are generally not known when a request first enters the system. In this paper, we highlight the fundamental challenges of regulating load and scheduling in SOAs while meeting end-to-end performance objectives on metrics of concern to both tenants and operators. We design Wisp, a framework for building SOAs that transparently adapts rate limiters and request schedulers systemwide according to operator policies to satisfy end-to-end goals while responding to changing system conditions. In evaluations against production as well as synthetic workloads, Wisp successfully enforces a range of end-to-end performance objectives, such as reducing average latencies, meeting deadlines, providing fairness and isolation, and avoiding system overload.

A Survey and Classification of Software-Defined Storage Systems

“…Differently, several systems have been proposed to address specific problems of cloud storage in a SDS-fashion (e.g., performance isolation, tail latency, resource fairness), which are well-suited to be further integrated as a submodule, storage functionality, or control algorithm of full-fledged SDS systems [74,82,98,104,124,125]. Pisces [98] introduced system-wide performance isolation and fair resource allocation at the service level in multi-tenant cloud environments, through a physically centralized controller and a twofold data plane setup.…”

Self-managed services using MirageOS unikernels

“…Building on the successes of distributed tracing in addressing resource management tasks for cloud applications [12,19,21], I propose that unikernels extended with appropriate distributed tracing capabilities represent a promising building block to address the challenges of developing and deploying finer-grain and larger-scale service-based systems that selfadapt to changes in workload and operation environment. This unique combination will allow to build auto-scaling management systems that are both generic and applicationaware as well as embodied within the service system.…”

“…when a QoS guarantee is about to be violated). Additionally, the request-centered approach would be useful to avoid unncessary scaling decisions due to artificial bottlenecks in upstream services caused by overloaded downstream services [19,21].…”

Quality-Elasticity: Improved Resource Utilization, Throughput, and Response Times Via Adjusting Output Quality to Current Operating Conditions

“…Suresh et al targeted the problem of collaboratively meeting deadlines in a service-oriented architecture [21]. Their approach hinges on rate limiting and scheduling on a per-service basis to optimize overall ability to meet deadlines.…”