Browse Prior Art Database

Improving File Retrieval Performance in Content Manager

IP.com Disclosure Number: IPCOM000031108D
Original Publication Date: 2004-Sep-10
Included in the Prior Art Database: 2004-Sep-10
Document File: 1 page(s) / 28K

Publishing Venue

IBM

Abstract

IBM DB2 Content Manager(CM) provides a common interface for managing content. A typical CM configuration includes a library and one or more resource managers. System Administrator can assign a user's default Resource Manager(RM) as a LAN cache server. A LAN cache server store files from other RMs in its caching area(staging area). When there is a cache hit, the copy in the caching area will be sent to client. When there is a cache miss, the LAN cache server has to get the file from the owning RM. Currently, we only cache the files that client previously retrieve. In order to increase the number of files that can be retrieved from a LAN cache server without increase the size of the caching area(which is limited by the disk space pre-allocated), we send out a valid replica if it is found in the LAN cache server. By doing this, we can improve the file retrieval performance in Content Manager.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 60% of the total text.

Page 1 of 1

Improving File Retrieval Performance in Content Manager

To increase objects availability, the CM Resource Manager (RM) Server can be configured to store replica of objects owned by other Resource Managers in its own pre-allocated permanent location. When the owning RM is offline, objects can still be retrieved from the replicating RM.

To improve the objects retrieve performance over the network, a RM can be configured to be a LAN Cache server. Remote objects requested will be retrieved from the remote RM and cached in the local caching area. Since the space of the caching area is limited, objects stored in the caching area but not referenced recreantly will be purged when the caching area reach a certain pre-defined threshold, for example, 85% full.

A RM can be configured to serve as a LAN Cache server and a Replicating Server at the same time. While retrieve requests will trigger the LAN Cache server to pull objects from other RM to its caching area, the replication orders executed in other RMs will push more remote objects to this RM server.

This dual roles configuration gives the CM an advantage to increase the cache hit ration. Replicas can be treated as pre-fetch objects waiting to be retrieved. Include the replicas when the LAN Cache server RM is searching object for retrieve, the RM virtually extend the caching area space, and increase the possibility of cache hit.

When a retrieve order comes in, if the requested object is NOT found in the caching area, an...