HDS has run a survey and found that fire, flood and human error account for most data loss. Tapes can be unreliable and the fastest and possibly safest way of recovering lost data is from replicated data held at a remote site.
Bob Plumridge, HDS' software director EMEA, says that as replication has grown in popularity a changed approach has paid host dividends. As an aside he notes that stored data needs are growing particularly because of compliance. Morgan Stanley is threatened with a $1.45 billion fine in the USA for failing to respond to a compliance request for e-mails. Subsequently they were found on tape and the authorities pounced. If they had been stored on disk then they would have been found. If that disk crashed but its contents had been replicated then they would still have been found.
HDS' TrueCopy sends data off site asynchronously. Once the base data has been copied than only updates are sent across the WAN to the target site. It's a 1:1 facility, primary array to secondary array. Updates to data on the primary are stored in a so-called 'side file'. These are time-stamped and held in cache. When it's time to replicate then the primary array sends cache contents to the secondary array in a 'push' fashion. There they are applied in the right time order to the base data to bring it up to date. In a high workload environment this can be a problem because the primary's host resources are being used intensively.
A new HUR - HDS Universal Replicator - was announced with TagmaStore. Plumridge says: "With Universal Replicator we changed to a journalling system. Journals are held on disk rather than in cache - which is finite. The journal is sent to the remote array and it pulls data over from the primary. This releases resources for the application workload on the primary array."
"This can be done for several different locations." It is a 1-to-many replication scheme, making things more secure. The interval at which the journal file is sent over the WAN is tunable at the volume level. You can choose to replicate critical volumes only and have the interval reflect the write intensity level of the volume workload. This means there's no need to replicate volumes holding paging data or static files holding system information or resource files.
Archiving and remote offices
For archiving to tape HDS can offer CommVault software which can run on a server local to an HDS array (or HDS managed array) and backup up array contents to tape. This opens up the possibility of placing mid-range HDS arrays, such as the newly-launched AMS and WMS products, in remote and branch offices.
HUR is run in these remote offices to replicate array contents to a central HDS data centre facility. Thus remote office data can be recovered from the data centre and no local backups need be run. Secondly the CommVault software can be used to write this data to tape and thus archive it, providing a two-tiered data protection facility from HDS. The boundary in the software space between HDS' data protection activities and those of the backup software vendors is changing.