Vitaliy S. wrote:Alex, can you please tell us what issues you're experiencing with this storage? Are they performance related? What is your configuration?
My colleague has a ticket open (with Veeam) about this, and it has been an ongoing issue. We have two problems with the appliance. Firstly, that we're not achieving the dedupe ratio that is advertised (We get 4:1 at best, they advertise 10-15:1). We have 72TB RAW, and we have around 300TB logical data stored. The appliance is now showing 15TB free raw disk.
Secondly, performance of the appliance is poor. We have a fully populated unit (36x3TB). We've been told by Dell support that in the 12 or so months since it was put in place, the "cleaner" has never completed
a cycle. We've now had it running with no other I/O for 9 days, and while it has reclaimed ~ 15TB of raw space, its not realistic to expect this to happen on a regular basis.
Sequential write seems to peak at ~ 100-150MB/s, so running more than one or two active full backups doesn't work very well. Running the cleaner 24x7 also kills write speeds. Running Tape backups with the dedupe appliance as the source also shows significantly slower performance than raw disk (30-50MB/s). I should mention - my preliminary testing shows this is the same when reading or writing directly to the appliance with Windows.
Our local Dell sales team are looking at the issue with us, but they're very keen to point at Veeam as the culprit. They also seem to think that nobody (apart from us) is using the DR4100 with Veeam, despite it being supported by Dell. The DR4100 integration guide suggests we use Veeam dedupe, with compression disabled, and apparently this is why we're only getting 4:1 dedupe.
I'd be interested to know if DD users are getting better dedupe ratios, especially when they have Veeam doing inline dedupe.
Vitaliy - Is there any way to scan files on disk for their Dedupe/Compression ratio as seen by Veeam? Apparently, despite Local Target Dedupe being enabled, we're still only seenig 1.0x or perhaps 1.1x dedupe displayed in Veeam.
If we could get some assistance in determining whether Veeam is in fact the culprit (at least for the dedupe ratio) that may assist.