FAST CLOUD: Pushing the Envelope on Delay Performance of Cloud Storage With Coding

Abstract

Distributed storage systems often employ erasure codes to achieve high data reliability while attaining space efficiency. Such storage systems are known to be susceptible to long tails in response time. It has been shown that in modern online applications such as Bing, Facebook, and Amazon, the long tail of latency is of particular concern, with 99.9th percentile response times that are orders of magnitude worse than the mean. Taming tail latency is very challenging in erasure-coded storage systems since quantify tail latency (i.e., xth-percentile latency for arbitrary x ∈ [0, 1]) has been a long-standing open problem. In this paper, we propose a mathematical model to quantify tail index of service latency for arbitrary erasure-coded storage systems, by characterizing the asymptotic behavior of latency distribution tails. When file size has a heavy tailed distribution, we find tail index, defined as the exponent at which latency tail probability diminishes to zero, in closed-form, and further show that a family of probabilistic scheduling algorithms are (asymptotically) optimal since they are able to achieve the exact tail index.

Simplex Queues for Hot-Data Download

“…where E[V j ] and E[V 2 j ] are given in (19), and f j is the limiting probability that an arbitrary job makes type-j service start.…”

“…where E[V j ] and E[V 2 j ] for j ∈ {0, 1, 2} follow from (19). We next estimate f j under high-tra c. Consider a join queue at the tail of the system, in which the tasks that nish service wait for their siblings.…”

“…$15.00 DOI: 10.1145/nnnnnnn.nnnnnnn simultaneously download from di erent replicas, or each user can issue simultaneous requests to multiple replicas and wait for the rst download to nish [6,7,15]. Data access systems arising in erasure coded storage have received a lot of a ention for all-data (see for example [12,14,19,24] and references therein) and some for hot-data download [16,17], and usually involve single or multiple inter-dependent fork-join queueing systems.…”

Joint latency and cost optimization for erasurecoded data center storage

“…A queuing model closely related to erasure-coded storage is the fork-join queue [15] which has been extensively studied in the literature. Recently, in [2], the authors proposed a heuristic transmission scheme using this Fork-join queuing model where a file request is forked to all n storage nodes that host the file chunks, and it exits the system when any k chunks are processed to dynamically tuning coding parameters to improve latency performance. In [4] the authors proposed a self-adaptive strategy which can dynamically adjusting chunk size and number of redundancy requests according to dynamic workload status in erasurecoded storage systems to minimize queuing delay in forkjoin queues.…”

On scheduling redundant requests with cancellation overheads

“…As a result, researchers have proposed several approaches to achieve better latency performance while living with this high variability. One of the most promising approaches is that of scheduling redundant requests to multiple components or servers [1]- [5]. That overheads.…”