ExaGrid Detailed Product Description

19

RAID6 Internal Storage with Consistency Checking

All ExaGrid internal storage is accessed using an industry-leading PCI RAID controller at the RAID6 level of disk

protection with a global “hot spare” disk. Since RAID6 keeps stripe parity on two disks, each ExaGrid appliance

can tolerate the loss of up to two disk drives at the same time. The first lost disk drive will initiate a parity

rebuild operation using the global hot spare as well as informing the backup administrator and (optionally)

ExaGrid customer support of the failure. A replacement disk drive is dispatched quickly, typically allowing

replacement of the failed disk the next business day. Loss of second disk does not result in loss of data since

the remaining parity disk allows for data regeneration; this extends even longer the time available to replace

the failed disk(s).

During normal operation, the RAID controller does consistency checking of the data on its disks in the

background, correcting any disk media errors using the parity disks.

Flash-Backed RAID Cache

The industry-leading PCI RAID controller has onboard writeback cache backed up by a super-cap powered

flash memory. Unexpected loss of appliance power does not result in backup data loss because any in-process

writes to any disk are preserved until power is restored.

Backup Data Checksums with Automatic Repair

As backup data is deduplicated, checksums are added to the deduplicated data as it is placed into the internal

storage area, called the “repository.”These end-to-end checksums cover the deduplicated backup data itself, and

are used to verify the backup data during processing and as it is read from disk. The deduplicated backup data

can optionally be replicated to a remote site; these checksums are used to validate the replicated data as well.

The ExaGrid software continually scrubs the repository data, confirming checksums and automatically

repairing any deduplicated data that does not match its checksum using data from remote site(s). This

automatic repair of deduplicated data is covered by one of ExaGrid’s patents.

Deduplicated Metadata Transactional Consistency

Metadata that tracks all of the deduplicated data is kept in a database and on internal storage. Software

techniques are used to ensure transactional integrity of all metadata changes, including flushing filesystem

pages into the flash-backed RAID onboard cache. The data flow of deduplicated backup data is protected

end-to-end by the combination of checksums (above) and metadata transactional consistency.

Internal Database Backups and Self-Describing Metadata

The database used to keep metadata that tracks deduplicated data is periodically dumped to internal

storage. These dumps are used to quickly restore the metadata database in the case of massive failure. The

database dumps are used as an optimization; the metadata kept on disk is self-describing and can be used to

completely rebuild the deduplicated data in the internal repository both at the local and remote ExaGrid sites.

Logging Filesystem

Backup data is kept in the ExaGrid internal storage on an industry-standard logging filesystem where file

activity is logged for integrity and quick repair after an unclean shutdown.