Interposed proportional sharing for a storage service utility

Abstract

This paper develops and evaluates new share-based scheduling algorithms for differentiated service quality in network services, such as network storage servers. This form of resource control makes it possible to share a server among multiple request flows with probabilistic assurance that each flow receives a specified minimum share of a server's capacity to serve requests. This assurance is important for safe outsourcing of services to shared utilities such as Storage Service Providers.Our approach interposes share-based request dispatching on the network path between the server and its clients. Two new scheduling algorithms are designed to run within an intermediary (e.g., a network switch), where they enforce fair sharing by throttling request flows and reordering requests; these algorithms are adaptations of Start-time Fair Queuing (SFQ) for servers with a configurable degree of internal concurrency. A third algorithm, Request Windows (RW), bounds the outstanding requests for each flow independently; it is amenable to a decentralized implementation, but may restrict concurrency under light load. The analysis and experimental results show that these new algorithms can enforce shares effectively when the shares are not saturated, and that they provide acceptable performance isolation under saturation. Although the evaluation uses a storage service as an example, interposed request scheduling is non-intrusive and views the server as a black box, so it is useful for complex services with no internal support for differentiated service quality.

Mitigating Performance Unpredictability in Heterogeneous Clouds

“…Many scheduling algorithms use relative values [29] to allocate resources [1], [2], [3]. [9] studied IOPS reservation in a IaaS.…”

An Adaptive Mechanism for Fair Sharing of Storage Resources

“…To address the similar problem in the domain of storage, various adaptations of these algorithms have been proposed. For example, YFQ [11], SFQ(D) [5] are based on Start-time Fair Queue (SFQ) [12]; SLEDS [13] and SARC [14] are adaptations of the leaky bucket algorithm; CVC [15] [4] adopts VirtualClock [7]. However, directly applying fair queuing-based algorithm to storage system can introduce unfairness within a short period of time (Section II-C).…”

“…An extension of SFQ, called SFQ(D) [5], maintains a queue for servicing up to D concurrent requests. With this extension, during a busy period, the virtual time is set to be the maximum start tag of requests in service.…”

“…To overcome this problem, we propose a novel algorithm as an extension of SFQ(D) [5]. Our new algorithm provides fair resources sharing defined by SLA for competing users with automatic adaptation to performance variation of diskbased storage systems.…”

Achieving Class-Based QoS for Transactional Workloads

“…Finding a good lower bound is a difficult problem (and is left as an open problem even in recent publications [13]). Our basic approach is to use a control loop augmented with queueing theoretic guidance.…”