OpenAFS logo
AFS & Kerberos Best Practices Workshop
Ann Arbor, Michigan
June 12-16, 2006
Home
List of Participants Agenda

Program

Tutorials and Workshop take place in at University of Michigan.

Thursday 15 June

Speaker Title
Kristen Webb, Teradactyl LLC Multi-Level Synthetic Backup Consolidation
This paper discusses the advancements of synthetic backup consolidation to meet the advantages of traditional level backup strategies. Synthetic backup consolidation has been generally unavailable for AFS. In most backup products with synthetic backup consolidation the process is limited to incremental and synthetic full backup processing. Extending these capabilities to multiple levels provides substantial benefits to scale backup systems while synthetic consolidation reduces the impact of backup operations on the networks and target systems. Altering level backup strategies and tape retention policies can provide control of granularity of data revisions in the backup system and greater control of storage costs. Multi-level synthetic backup consolidation provides many types of backup processing for AFS including network full and true incremental, network-synthetic cumulative, synthetic cumulative, synthetic partial-cumulative, and synthetic full backups. The use of backup server disk caching provides for fast, parallel processing of backups from multiple AFS servers and high data transfer and compression rates to modern tape devices. Supplemental disk library storage enhances synthetic backup and restore processing to further increase performance versus tape storage alone. These capabilities allow large AFS installations to consolidate backups from tens of thousands of AFS volumes and tens of terabytes of data to a few dedicated, centralized backup servers. This minimizes the impact of the backup function on AFS cells and networks, and further reduces backup hardware and storage costs. Additional topics specific to AFS include volume tracking, file cataloging, and vice partition recovery.

 

 

University of Michigan