Browse Prior Art Database

Workload optimized multi tenant cluster file system for cloud applications Disclosure Number: IPCOM000210139D
Publication Date: 2011-Aug-26
Document File: 1 page(s) / 126K

Publishing Venue

The Prior Art Database


Archival and enterprise content management solutions in the cloud are typically priced on GB of customer data per month - hence they are very storage cost sensitive. Offering a scale out - multi tenant capable storage subsystem is essential for cloud services to be competitive. Traditionally different technologies such as SAN and NAS have been used for storing archived objects and high IOPS objects such as full text indexes and databases. As these multi tier models do not scale well in the cloud, are usually not multi tenant capable and very costly and new approach is needed for Multi Tenant Cloud Services. The approach described uses a Cluster file system such as IBM GPFS and wide striping to meet all requirements at much lower complexity and price point.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 100% of the total text.

Page 01 of 1

Workload optimized multi tenant cluster file system for cloud applications

The idea is a design for a workload optimized usage of cluster file systems (i.e. IBM GPFS). On top of the workload optimization the concept of file sets is used to isolate tenant data - this allows for a very cost effective model of just a very small number of file systems while still being able to address the data isolation requirement of multi tenancy. One very important aspect to reduce the costs is the usage of only one kind of storage model and tier: very high density slower rotation disk drives (e.g. 2 TB SATA disks). The essential design point is that most of the capacity (i.e. 80%) is used for dormant archived data and that about 20% of each spindle is used for high performance data. Using this concept with wide striping of a cluster file system ensures cheap storage and yet high performance because all data is striped across all disks. If more performance or more capacity is needed one simply needs to add more disks to meet both requirements.

(This page contains 00 pictures or other non-text object)