Data Storage in Cloud Based Real-Time Environments

Data Storage in Cloud Based Real-Time Environments

Sai Narasimhamurthy (Xyratex, UK), Malcolm Muggeridge (Xyratex, UK), Stefan Waldschmidt (Digital Film Technology, Germany), Fabio Checconi (Scuola Superiore Sant’Anna, Italy) and Tommaso Cucinotta (Scuola Superiore Sant’Anna, Italy)
Copyright: © 2012 |Pages: 23
DOI: 10.4018/978-1-60960-827-9.ch013
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

The service oriented infrastructures for real-time applications (“real-time clouds1”) pose certain unique challenges for the data storage subsystem, which indeed is the “last mile” for all data accesses. Data storage subsystems typically used in regular enterprise environments have many limitations which impedes direct applicability for such clouds, particularly in their ability to provide Quality of Service (QoS) for applications. Provision of QoS within storage is possible through a deeper understanding of the behaviour of the storage system under a variety of conditions dictated by the application and the network infrastructure. We intend to arrive at a QoS mechanism for data storage keeping in view the important parameters that come into play for the storage subsystem in a soft real-time cloud environment.
Chapter Preview
Top

Introduction

The cloud environment mandates multiple “tiers” of functionality before the applications eventually get access to storage. These tiers bring about a disconnect between application requirements and storage functionality. The problems of providing a predictable and reliable storage subsystem is further exacerbated when real-time applications need to be serviced by the cloud.

The storage subsystem for distributed real-time applications within clouds requires the following storage optimization characteristics:

  • a.

    Capability of dealing with rapidly fluctuating demand from hundreds of applications accessing storage

  • b.

    Ability to leverage commodity storage subsystems to increase the Return on Investment and reduce costs for storage service providers within the cloud

  • c.

    Scalability to accommodate an ever increasing number of applications

  • d.

    Use of readily available, mostly open source software components to implement storage

  • e.

    End user performance that does not get affected with constant increases in storage capacity as noted in (c)

  • f.

    More flexibility of provisioning and managing storage than currently available in standard enterprise storage implementations

Storage QoS which is fundamentally absent in enterprise storage gains significance in real-time application based clouds. QoS defined at various tiers of real-time clouds percolates to storage specific Service Level Agreements (SLAs), which needs to be reliably adhered to.

This chaper is organised through the following sections:

  • Storage Optimization: This section discusses current day storage subsystems optimized for soft real-time environments through an understanding disk drive technologies, Storage virtualization technologies and file system technologies. The chapter also discusses the state of the art on research in Storage QoS. The arguments lead in to the use of Lustre™2 based storage solution with the QoS innovations we provide.

  • Storage Platform Processing: This section discusses a Lustre file system based storage platform and how it fits in to the overall soft real-time cloud environment framework.

  • Providing QoS: This section discusses our Storage QoS model implemented for our storage platform.

  • Conclusions and future work: Concludes the chapter highlighting next steps.

Complete Chapter List

Search this Book:
Reset