The Battle of the species: Veeam Native vs. EMC Data Domain

[skydok]

Hi Guys,
As part of my travels between customers i'm always getting the "Why should i invest in a Deduplication appliance?"
Well it has it advantages of course but lots of customer say they preffer to invest in 6TB Drives in a JBOD instead to get the same capacity at the end.
THEY BELIVE IN THE VEEAM NATIVE Compaction & LOW-COST DRIVES against the technology of deduplication appliance.

WELLLL...I wanted to once and for all show them the numbers so i did a DEMO for them and now you can see the results also.

http://skydok.com/2015/08/09/the-battle ... storeonce/

ENJOY!

[chrisdearden]

Can you repeat the test but with non duplicate data for the incremental workloads ?

[skydok]

There will be Volume #2 Update for the post which will include the follwoing:
1.Non-Duplicated Data
2.Enterprise Applications (Exchange,SQL,FS)

I belive i will come to this in the Next week

[chrisdearden]

Look forward to it

[dellock6]

nice job indeed! Only doubt is the ingest performances, as max speed seems the same for each job, as the bottleneck is source...

It would be also interesting to see "restore" performances, once in a while
There's obviously a trade-off in choosing a dedup appliance, and a restore example would better explain in some designs a non-dedupe storage is preferred (and also on ingest performances)

[skydok]

I need to save somthing for later , no ?!?
Of coure the restore comperission was also planned and will be in the next Post as an indepent subject.

[Gostev]

Good first step, but yes - I would love to see results on more real-world workloads:

1.Non-Duplicated Data
2.Enterprise Applications (Exchange,SQL,FS)
I believe i will come to this in the Next week

1. I think we can all guess the results here are you perhaps thinking just a bunch of JPEGs here or something?
2. Now this will be a fun test - as contrary to popular belief, these workloads do not dedupe well at all

[skydok]

1.I'm going to Mix Pics/Movie Clips and Bin files. Any other suggestion?
2.I'm Sure it will be intresting too Do you have any suggestions here as weel?

I like that you all started to get into the details i deffently now do this in the comming weeks and will publish the Updates.
You will be notified don't worry .

[petergrande]

Great comparison post on dedupe factor differences , if I may give you a suggestion since I can see that source is probably the limiting part here, not sure what it is that limits it, Disk, NIC and so on.. anyway if you can bump up the source speed in any way that would be great for the performance comparison maybe, we use DD990s and I`ve seeen transfer speeds up around 3000 MB/s source deduped from a single host (not with Veeam though, we haven´t moved our Veeam environment to DD yet), also remember DD can handle quite a lot of simultaneous sessions with good results, the max sessions depends a little on which DD model though, so if you run a larger workload with more VMs simultaneously would also be great to see, not sure here if the CPU Core limitations apply to Veeam B&R in combination with DD, but I guess it does..

[aporter]

Thanks for running these tests and for sharing the results. It would be interesting to see how a target running Windows Server 2012 deduplication stacks up with the same tests as well.

[mairiesuresnes]

Very useful! Thanks

[skydok]

I would deffently like to pumb up my source , but currently the hardware is what it is I'm Testing currently the same Scenarios on physical HW as well (Physical DD etc), it will take my two-three weeks until all test are done . Transfer Speed isn't somthing i was looking to compare ,more the evffectifve ness of the capacity savings. Therfore all speed mettrics are simillar there which i ignored for now.

Great comparison post on dedupe factor differences , if I may give you a suggestion since I can see that source is probably the limiting part here, not sure what it is that limits it, Disk, NIC and so on.. anyway if you can bump up the source speed in any way that would be great for the performance comparison maybe, we use DD990s and I`ve seeen transfer speeds up around 3000 MB/s source deduped from a single host (not with Veeam though, we haven´t moved our Veeam environment to DD yet), also remember DD can handle quite a lot of simultaneous sessions with good results, the max sessions depends a little on which DD model though, so if you run a larger workload with more VMs simultaneously would also be great to see, not sure here if the CPU Core limitations apply to Veeam B&R in combination with DD, but I guess it does..

[skydok]

Nice Idea , I'll add it also as a Feature request

Thanks for running these tests and for sharing the results. It would be interesting to see how a target running Windows Server 2012 deduplication stacks up with the same tests as well.

[lightsout]

Very nice! I'd like to see what happens if you enabled Windows' host based deduplication on top of it.

[skydok]

Its a Secret

Very nice! I'd like to see what happens if you enabled Windows' host based deduplication on top of it.

[ian0x0r]

Nice tests and good comparison.

Would be good to see your results as well for GFS retention criteria. This for me at least is where Dedupe appliances such as the DD come into their own as a secondary storage target for longer term retention (backup copy jobs). Would be good to compare that against Server 2012 dedupe though as already requested . For recovery speeds though, JBOD / Veeam Native is far quicker in my experience.

Ian

[ar63]

What Veeam Job compression level is better with a data domain storage? In my jobs i put "none" in that field bacause i prefer that is DD make dedu. But is it really the better choise?

Thanks.

[skydok]

The DD is first of all using the DDboost mechanism which is very good for less network throughput.
The DD reduction will be very very good after that so i would recommend it for you. Don't use veeam's side therefore you have the option of Dedplication appliance as Repository.

[HannesK]

It makes no difference

As long as you do not have a gateway server in front of your DD it is irrelevant which compression level you use as this setting is ignored if proxy and repository are on the same server. Compression can save bandwidth if you have a gateway server but it has no influence on DD data reduction in any case. Reason:

If you add a DD as repository it will automatically be configured with the "decompress backup data blocks before storing" setting