HP officially joined the data deduplicators club today after several alert storage news sites including our own Byte and Switch broke the news from a premature update of HP's Turkish website. As expected they're adding Sepaton's DeltaStor, which they're calling Accelerated DeDuplication, to the VLS VTLs they've been OEMing from Sepaton for the past few years. More interesting are the new D2D2500 and D2D4000 appliances HP is targeting at what if you're HP or EMC is the SMB market but to those of us without billions of dollars of sales a year would call midmarket. Amazingly for HP the D2D2500 sets a new low for a data deduplicating target at just $6500 with 2.25TB of usable disk space after the RAID 5 overhead.These boxes use HP's own inline "Dynamic Deduplication" using hash based technology developed at HP Labs. The 1u D2D2500 emulates 1 to 6 independent eight slot autoloaders over a 1 GB/s iSCSI link while the 2u D2D4000, which is available in models with 3TB and 7.5TB of usable space in a RAID 6 array, can also emulate an HP MSL tape library and connect to your SAN via Fibre Channel. D2D4000 models range from $19,000 to $43,000.
You can also connect an HP LTO tape drive to either box and export or copy data from a virtual tape to a physical tape without running it through the backup server and software a second time. I've only seen this feature before on enterprise VTLs based on the FalconStor software.
While I applaud HP for bringing the cost of deduplication down to the masses I think they made a couple of mistakes with these devices. First I'd rather see a CIFS interface than the autoloader emulation even if it means giving up the direct tape export function. Almost all backup programs today can use a NAS device as a backup target but adding a SAN attached autoloader could be an adventure for many in the target market not to mention that Symantec charges $500-750 additional for the Backup Exec San Storage Option and again for each tape drive through the library expansion option. On the plus side data on a VTL is less vulnerable to malware on the backup server wiping it out.
I also wish the D2D2500 would emulate a library not a group of autoloaders. Having separate pools of virtual tapes will be a problem some day as emergency restore requests come for 2 sets of data that belong to the same autoloader so they have to be processed sequentially.
HP promises replication in a future firmware release that could arive as soon as the end of the year. Maybe a CIFS/NFS interface could be delivered then too.