Data Domain with DDBoost - Source Disk I/O Reduction?

Question

Hello All, I'm currently in the market for a disk based backup target and being an EMC Networker customer and an EMC Storage customer it is obvious to at a look at Data Domain. There are a number of viable enterprise class vendors in this space but some of the features that particularly set Data Domain apart from competitors for us - were specifically the integrations within Networker from a workflow perspective and DD-Boost. My question here is specific to anyone who has had experience with the recent iterations of DataDomain appliances with DD-Boost in a client-direct model (ClientServer --> DD appliance). The way DD-Boost was explained by Sales and Engineering was that in this scenario that DD-Boost would run on the client server and 'with just a bit of CPU overhead' reduce data-over-wire and reduce time-windows even further. There is no doubt that less data to send to the target is certainly a time-savings but I'm curious to know if DD-Boost is purely a network I/O saver or if there are benefits to the amount of IOPS hitting the source disk. We're not necessarily network-IO constrained but certain data-sets are certainly disk-IO constrained. • Does anyone know if DD-Boost aides / has a real benefit to disk-IO reduction on the data source during backups?• If so - what sort of % or rather, are you able to explain a bit as to how it may accomplish this (trying to step it out in my head)• Looking for any thoughts / experiences the community may have using DataDomain in general with Networker -- should I look at other vendors why/why not? Thanks in advance also if this is in the wrong spot or against the normal format please let me know

ble1 · Answer

Even with partial (or full) source de-dupe - data has to be read on client so I think nothing really changes there.  I think advantage as you explained might be visible where storage array might push data to DD - in solutions like ProtectPoint (which are still not integrated with NW nor officially out there yet).  As for DD vs others, I still think DD is leading the pack per se and if you have components which integrate - it is no brainer where to go.

crazyrov · Answer

DDBoost + client direct means part of the de-dupe process is completed at the source. To do this the client machine still has to read through the entire saveset to achieve this.

I have used only DD for now, a couple of HP StoreOnce but that again is not tightly integrated with the backup applications

NetWorker

Was this post helpful?