Gartner mages throw deduping backup appliance bones, claim EMC's in lead

Re: Where is Oracle/Sun/StorageTek? CommVault?

This MQ is for target dedupe devices only

Re: Where is Oracle/Sun/StorageTek? CommVault?

zfs dedupe is in a different league. It's more for online storage not for backups. I am in the midst of migrating from a ZFS (Nexenta) based dedupe+compression store to HP StoreOnce, ZFS gives me less than 6:1 dedupe(on a data volume dedicated for backups).

StoreOnce on my biggest file set is over 40:1 and still going (I suspect will continue to grow as I expanded the # of backups to store from 14 to 60 and I'm not even at 30 at the moment). My biggest backups are from our OLTP mysql instance - StoreOnce says there is 10.9TB of user data stored in 241GB of disk space (less size than one full pre-dedupe backup!). All other backups combined are less than 1.5TB of user data written.

Wherever possible (mysql especially) we are doing full unencrypted, uncompressed backups to get maximum dedupe (same as we were doing on ZFS). Percona xtrabackup is our backup tool of choice for MySQL at the moment.

By contrast devices like HP StoreOnce(or data domain etc) are REALLY REALLY bad for online transaction type stuff(they urge you in their docs to not attempt it). It's really built for sequential operations.

I only use StoreOnce in NAS mode, so can't vouch for any other abilities (an update in July increased it's NAS scalability to 1 million files per share from 25k which was too limiting for us - 1 million is plenty though). The one thing that StoreOnce can't do that I'd like it to do is support for symbolic (or hard) links on the NAS. I exchanged some emails with the HP product manager for StoreOnce a few months ago and they were going to look at whether or not they could add that feature.

I'm not personally aware of a dedupe Disk to disk backup appliance that is similar to StoreOnce or EMC Data Domain from Oracle or Commvault (though I haven't looked at either recently).

Yes but Dell's ability to execute is limited by their relatively low end product shoe horning Ocarina into a backup appliance, IBM's is that they lost focus and fell hopelessly behind with the Dilligent product. Exagrid is post process and although it scales out, each node isn't particularly large and scale out is still limited to relatively small environments.

HDS Just bought Sepaton who despite being one of the first to market and being very scalable have had a number of directional changes and struggled for mind share as a result. With HDS in the driving seat they could get much more focus going forward but given their relative complexity I doubt they'll be able to take huge market share.

EMC with Data Domain have a good product and are way ahead of everyone else in terms of market share, ecosystem support and thus their ability to execute. Technically they look to be falling behind the likes of HP, but the backup market, especially SW and appliances has always been sticky, once you have footprint you're in for the long haul.

HP has the more scalable and arguably more functional solution these days. However they've traditionally struggled with the wider backup ecosystem support. That looks to be changing though as the StoreOnce product gets more focus and matures.

There are also plenty of software only solutions out there, but these don't tend to stack up at scale where some of the bigger appliances play and if you have a big appliance at the core, it typically makes sense to have the smaller appliances at the edge.