IBM rejects deduplication for new VTL
A 3.4PB beast
IBM has revealed a new TS7530 virtual tape system with no deduplication capabilities at all, just hardware compression.
The Virtualization Engine TS7530 is presented by IBM as an intervening system between open systems servers and tape with tape being the final resting place for data. It is designed to augment the tape backup and restore processes in large tape environments.
A virtual tape system presents a drive array to backup software as if it were an automated tape drive or library.
The TS7530 replaces the previous TS7520 and uses 1TB capacity SATA drives. Like its predecessor it uses FalconStor's very popular VTL software product.
It supports up to 4,096 virtual tape drives and 512 virtual tape libraries and works with Windows, Unix and Linux servers via Fibre Channel and iSCSI connections. Its maximum raw capacity is 1.7PB, or 3.4PB of compressed data.
Despite having bought deduplication supplier Diligent, and having the relationship with VTL software supplier Falconstor, which has its own SIR deduplication product, IBM has decided to rely only on hardware compression and limit users to a 2:1 compression ratio rather than the 20:1 or greater ratios obtained with de-duplication.
The rationale behind the rejection of deduplication might be that the TS7530 is not a stand-alone backup target but a way-station for data en-route to tape. As such it would, if it stored deduped data, have to reconstitute it before exporting it to real tape drives. But competing suppliers do ship deduping VTL products.
Sun, for example, uses FalconStor's SIR deduplication in its open systems virtual servers. Its X4500 'Thumper' hybrid server/storage system is the basis for its VTL-V product which uses the FalconStor dedupe SW and holds up to 720TB of deduped data (20:1 ratio) on its 36TB of disk capacity. Sun is developing the larger VTL-E with FalconStor. It also sells Diligent's (now IBM's) ProtecTIER dedupe product on an X4600 server and 6000 array offering 168TB of raw disk capacity.
Diligent's inline dedupe, carried out at backup data ingest time, is thought to be better suited to enterprise VTL applications than FalconStor's post-backup process dedupe.
It is curious that IBM did not choose to use either Diligent or FalconStor's software in the TS7530. Perhaps the processor is not powerful enough to run the Dilient ingest-time deduplication software and so backup times would lengthen. However, FalconStor, with its post-process design would not slow down backups, unless IBM thought that the TS7530 would be ingesting data pretty much all the time and would have no spare capacity for dedupe or data reconstitution work.
This would mean that IBM rejected deduplication because the TS7530 product does not have powerful enough processing resources to carry out deduplication or reconstitute deduped data as well as cope with incoming backups.
The TS7530 is priced from $251,000.
Copyright © 2008, Blocks & Files.com
Sponsored: Hyper-scale data management