Current implementations of backup processes are manual, labor intensive, and inundated with problems caused by human errors. Backups that fail can often go undetected leading to potential data loss following a failure. Tapes must be manually inserted, rotated, and removed for off-site transportation to ensure recoverability. Due to high expenses associated with manual administration, a case for automation can be made where robots are used to improve tape management. Centralized backups group tapes together by pools whereby a number of backups can be multiplexed onto a tape. As a tape is filled the backup continues using another free tape within the pool which significantly enhances manageability. A reduction in the number of tape drives to be managed leads to cost reduction.
Backups require an increase in application server activity in order to fetch data from disk and write to tape. Application servers are usually busy processing large volumes of latency and performance sensitive data. The extra CPU cycles consumed for data movement and scheduling while conducting backups can often prove costly to the application itself. A server-free solution is targeted at eliminating the performance impact of a backup when performed by the application server itself. This architecture migrates the data-mover and connection-broker roles to a dedicated backup server. A robust backup solution must address the following concerns:
1.Optimize use of backup resources including tapes, drives, and operating time
2.Minimize impact to application traffic
3.Segregate backup domains to reduce CPU overhead on application servers
The following section provides details on most common implementations of backup solutions.
LAN-Free Backup
LAN-ree backup enables each application server to move data directly to a tape device over the storage network without going through a dedicated backup server. Using a commonly known shared-storage option, each application server acts as a media server in that it moves backup data directly to tape. Each server then arbitrates for a tape drive and reserves that tape drive during the backup process. Application servers also can be configured using dedicated access to tape drives within the tape library unit instead of the shared option. The flow of data through the storage network allows for a reduction in LAN traffic. A tape may also be shared amongst applications with multiple backup streams multiplexed onto managed tape libraries and drives. The LAN may still be used to pass metadata, context tables that track location of changes in data, back and forth between the backup server and the client but actual backed-up data is passed over the storage network. A typical LAN-free implementation is depicted in the
figure below.
LAN-free backup segregates data and backup domains but doesn’t alleviate CPU load on the application servers that still must fetch backup data from disk. Server-less backup as described below addresses this concern.
Cisco Systems, Inc.
All contents are Copyright © 1992–2002 Cisco Systems, Inc. All rights reserved. Important Notices and Privacy Statement.
Page 3 of 11
