Resolution-aware network coded storage

TL;DR

This paper demonstrates how coding strategies in storage area networks can significantly improve quality of service metrics like blocking and saturation probabilities without extra storage, by leveraging resolution-aware coding schemes.

Contribution

It introduces resolution-aware coding schemes for SANs, including network coded storage and resolution-aware storage, to enhance quality of service under bandwidth constraints.

Findings

SR NCS reduces blocking probability by an order of magnitude.

MR URS significantly improves saturation probability over non-resolution systems.

CRS further enhances saturation probability savings over URS.

Abstract

In this paper, we show that coding can be used in storage area networks (SANs) to improve various quality of service metrics under normal SAN operating conditions, without requiring additional storage space. For our analysis, we develop a model which captures modern characteristics such as constrained I/O access bandwidth limitations. Using this model, we consider two important cases: single-resolution (SR) and multi-resolution (MR) systems. For SR systems, we use blocking probability as the quality of service metric and propose the network coded storage (NCS) scheme as a way to reduce blocking probability. The NCS scheme codes across file chunks in time, exploiting file striping and file duplication. Under our assumptions, we illustrate cases where SR NCS provides an order of magnitude savings in blocking probability. For MR systems, we introduce saturation probability as a quality of service metric to manage multiple user types, and we propose the uncoded resolution- aware storage (URS) and coded resolution-aware storage (CRS) schemes as ways to reduce saturation probability. In MR URS, we align our MR layout strategy with traffic requirements. In MR CRS, we code videos across MR layers. Under our assumptions, we illustrate that URS can in some cases provide an order of magnitude gain in saturation probability over classic non-resolution aware systems. Further, we illustrate that CRS provides additional saturation probability savings over URS.