VMWare: NFS & Dedupe

Hi,

To get similar efficiency gains in your virtualized environments using your EMC gear, you can use a V-Series controller from NetApp. You would provision your LUNs from the EMC array to the V-Series controller, and it would treat them as your current NetApp systems do their disks. In fact, once the array LUNs are provisioned, it looks exactly like your existing NetApp systems as far as hosts are concerned. This would allow you to make use of the otherwise quality SAN array EMC has provided, while getting all of the benefits you are now seeing with NetApp (converged network, NFS, Dedupe, etc..) Ask your NetApp sales rep for more info.

http://www.netapp.com/us/products/storage-systems/v-series/

Thanks!

- Dan

Jim, thank you for picking EMC. 

I know one of my colleagues is reaching out to you to help you get the most out of your EMC Unified infrastructure in your VMware environment.

EMC (and for that matter all people who make storage arrays) have different ways of getting the most out of the least.   In our case, it's deduplication for general purpose NAS, compression for many other use cases, Virtual Provisioning, FAST automated tiering and FAST cache (and much much more).

Often, when switching from one platform to another (in your case, from NetApp to EMC), there's a short period of learning about how to get the most out of your new platform.  EMC Unified platforms are simple and easy - but are of course, different than what you've used in the past.

We're all looking forward to helping you get the most out of the least - and EMC guarantees it!

http://www.emc.com/products/unified-storage-guarantee/

Again, thank you for being a customer, and we all look forward to serving you.

P.S. Dan - glad to see NetApp people trolling EMC's Community.  I'm glad that we're wide open and transparent - it's good for customers, and good for our partners.   It's clear that in spite of our transparency, you're no expert on EMC, as we are surely not on NetApp.    I'm also glad that you feel like trolling is a good use of NetApp folks' time.   Thank you!

Chad

I appreciate responses from both sides. I've also opened tickets with both EMC & Netapp.  I want to ensure I'm doing everything I can to maximize dedupe and performance from both solutions.

Dan - there was no personal attack, only a statement that I'm glad we're open and transparent.   You didn't disclose that you worked for NetApp, it was a guess, but I'm glad it was right, otherwise I would feel really bad.

The customer is getting their responses from EMC on how to best leverage VMware NFS datastores on EMC Unified Storage platforms, and I certainly hope they share their experience.

My comment to you was, and still is, I'm glad that NetApp employees find trolling the EMC Community a useful way to spend their time.  I don't think that's a personal attack.  If you took it that way, I'm sorry.

Yes, I do disagree that putting a NetApp v-Series in front of a EMC storage array is the best way to increase a customer's $/GB, W/GB, sqft/GB, $/IOps, W/IOps, sqft/IOps efficiency.   It would certainly add cost, space, and of course, they would need to migrate all their data (lowering their utilization every step of the way) - in turn to turn on a feature (block-level dedupe) that NetApp does, and EMC currently does not.   In doing that process, they would lose things that EMC does, and NetApp does not.

Conversely, they could use the EMC Unified array to it's maximum, leverage FAST/FAST cache and very possibly get to better utilization.

There are other ways to be efficient than the way NetApp does it, and we'll help the customer get the most they can out of their EMC infrastructure.

I'm sure that as a Tech Marketing Manager for v-Series at Netapp, you would disagree with the above - but this isn't a marketing medium, it's here to help customers and partners.

I'd like to suggest that we let the process (the customer has has opened tickets, and we are engaging to make sure we help them the most we can).   In the end, I hope they share their experience.

Thank you, and happy holidays!

Hi Chad,

My apologies if you thought I was trolling.  I do work for NetApp, as Google will attest, but I may not always and forever.  So were I simply trolling, I don't think I'd use my real name.  I'll note that you chose a personal attack in lieu of a critique of my suggestion.

I am an EMC customer who manages his own arrays, by virtue of my role as V-Series expert.  So while I would not pretend to be conversant in all aspects of EMC's product offerings, I am aware that your own manuals dissuade people from using your Dedupe on workloads like virtualized servers or desktops.  You do have some efficiencies that enable space savings, but they simply are not on par with what NetApp offers for these workloads.  That's not to say that NetApp is always better for everything.  Just that it's better for Deduping virtualized server/desktop deployments.

As the OP asked how to get his EMC gear to work as well as his NetApp gear in this repsect, I offered the only possible way.  Mine was the best answer to his specific querey.  Do you dispute this?

- Dan

Thanks for the kind reply, Chad.

To be trolling, you must be a troll.  And tis no greater personal attack on a message board.  My response would be my response were I still a sysadmin. And no effort was made to hide what is public information concerning my current employer. 

I can't speak to Jeff's best options with respect to those metrics without much more information about his environment.  It is quite possible that the 50%-90% space savings he would see with NetApp may not justify the expense incurred.  Just because it usually does, doesn't mean it always will.  The scale of his deployment and age of his EMC array would greatly impact the ROI of a V-Series controller.  Those are things his NetApp account team, being more familiar with Jeff's storage issues, could help figure out. 

I don't dispute the possibility that leveraging EMC tools and features as you suggest would be sufficient to make up for not being able to dedupe without significant performance overhead.  I'll reserve judgement, of course, but I'm not incredulous.  I'm glad that you guys are focused on helping him.  I take for granted that helping our customers is everyone's foremost concern.  Hopefully Jeff is able to get a nice discount on his next purchase. 

And lest you think I'm some nasty partisan, I quite like my Clariions.  They do what I ask them to do well.  They are easy to manage, and when I've needed it, support was helpful.  Keep up the good work!

May your Holiday be equally joyful,

Dan

All,

                I thought this thread should have my results posted to it. We have done some extensive testing and had equipment from both vendors onsite. We read through many blogs, articles, white papers and manuals during testing. We used the same data sets and same VMware data stores for these tests.

Let’s start off with the technical specs:

1.)    We found the EMC had faster CPU and more Memory per host and cluster

2.)    Redundancy – The NetApp isn’t a true redundant system since the controllers are separate. As we saw NetApp was 4 9’s availability versus 5 9’s from EMC

3.)    PAM/CACHE – The NetApp cache is read only versus EMC cache as read/write

4.)    PAM/CACHE – When either controller reboots the CACHE is cleared versus moved to the other controller on EMC

Let’s discuss capacity now:

1.)    We used 1.3TB of RAW disk capacity

a.)    EMC had 81% of the RAW capacity as useable

b.)    NetApp was 48% of the RAW capacity as useable

2.)    Dedupe rates

            a.) EMC was a 46% dedupe rate plus we saw gain from compression

            b.) NetApp a was 90% dedupe rate

Results:

                Manageability was one of the major concerns in our organization as well.  NetApp is simple to manage while allowing a robust CLI and a great V-sphere plug-in. However filer view is a very old interface and needs some updating as well.  With EMC's current platform, we found that Unisphere is a simple to manage platform and the new vSphere plug-in is robust and works great.

One thing we did note was that response time on EMC fiber channel LUN’s was better than the NetApp LUN’s.  Now in our environment I’m not sure that would matter a whole lot but it was worth noting. Reading various blogs also indicated the same results on FC & iscsi; noting that EMC had better performance than NetApp.

I think there is no doubt that the WAFL file system is a great file system and block level dedupe beats out file/byte level dedupe. NetApp appears to have better response from a NFS mount perspective however we had some permission issues getting mounts to work on VMware which was seamless with the EMC.

All in all so far we couldn’t find a compelling reason to choose NetApp for our environment.  I think in the end the dedupe differences seem to even out in regards to useable capacity. Well that’s it from our testing and research, I hope this ends up being helpful to someone else out there.

Jeff, thank you for being an EMC customer.  I sincerely appreciate your followup to share your findings, and your thorough work.

If there's ever anything I, or any other EMCer can do to help - we're here for you.

Disclaimer: NetApp Employee

I attempted to reach Justin, and unfortunately he has not responded.  I've reviewed the data and would offer the data with the following observations:

1. The NetApp system is being deployed manually without following best practices and a such returns such poor usable capacity numbers. By simply using the NetApp VSC plug-in for vSphere the customer would be up and running with optimal storage designs in literally one minute.

I believe attempting to configure any technology while not following published best practices leads to less than steller results.

2. NetApp block level dedupe (both disk and cache) is much more advanced than any offerings from EMC on any of their platforms, both in terms of storage savings and performance.

It is unfortunate to learn that I this customer didn't have a world class experience. I'll take the blame for that; however, I'd challenge the assumption of the numbers being shared as what one should expect to see with their deployment.

I would add, both of the point I made above (1 & 2) have been validated by VMware performance engineering in TR-3856, which can be downloaded here:

http://www.netapp.com/us/library/technical-reports/tr-3856.html

Cheers,

Vaughn Stewart

NetApp

Director & Evangelist for Virtualization and Cloud Soluitons

Sorry for the late response! Thank you for the report, I'll review it.

I can tell you we did use the plug-in just as we did with EMC. We also used the same snap reserve on both sides. Plug-in's arent very helpful when it comes to exisiting datastores. I also would like to point out one point I did not mention and that is that Netapp does dedupe fiber channel LUN's due to block level dedupe. Taking this into account you do get some more space. We did get a document from our SE stating the 5 9's as well.

Great information from both sides, thanks for your assistance in our testing.

Jeff, Vaughn - thank you.  Vaughn - please, let's not use this post to post our own trumpet blowing.  The customer had a question, they had both our stuff, they used our plugins.

No one disputes that the filesystem->consumed space with NetApp block-level dedupe was more efficient than our current EMC compression and file-level dedupe with VMDK-centric datasets.

The customer's findings were that the end-to-end (raw to used) utilization was close, when using the existing GA capabilities of each which he had on site and was working with himself (using the vCenter plugins from each). 

Of course each of us could have brought more to bear (FAST Cache/FAST for EMC, FlashCache for NetApp as just one example of many).   It also doesn't dive into the pre-post VAAI testing findings.

I'll also note that the document you link to is a NetApp TR - authored by you Vaughn and other NetApp folks, which of course will have a certain message/bias, no? 

It is not a VMware document, and I would suggest it's not a good idea to represent it as such.   It's like the constant refrain which I hear over and over - of "VMware uses Netapp for _____ IT workload" (which they don't), or "NetApp is the NFS/iSCSI reference for VMware" (which it's not).  VMware supports us all - and we each have things to do which are great and help customers.

Of course, there's much, much more that EMC could do to help the customer, above and beyond what Jeff menioned.

But - same core comment as my earlier comment to the earlier NetApp posters.   The focus of this post is not to be a NetApp vs. EMC commercial, but rather help the customer get the most out of what they have, which we have done.

Again, thank you Jeff, and thanks again for posting your findings.

First an apology, I stated Justin when I meant Jeff.  Please forgive me.

The findings here are still unrealistic, as Jeff noted he had manually configured the testbed prior to using the NetApp plug-in.  I've stated this many times in the past, if Chad would like to compare technologies in an open format (we've discussed VDI, but it could be any cloud based solution), I'd be happy to sign off on NetApp's participation.

As for the TR-3856... as Chad knows, technology partners cannot publish performance results obtained with vSphere without the consent of VMware.  Legally I cannot share the make and model of the array we compared the NetApp to in the test (as the declined to aprove the results), I can assure you VMware performance engineering reviewed all results and approved the results published in the report.

As I stated in my previous message, I am truly sorry that Jeff did not have a world class experience during his evaluation of NetApp technology.  Jeff if you'd be so kind as to share your experience, I'd be in your debt. Feel free to contact me anytime via email at v@netapp.com.

Chad I expect more from you than to promote claims of poor usable storage as 'par for the course with NetApp' - you know very well the details of our usable capacity.  I guess that's why it's called competition.  take good care of Jeff (or I'lll earn his business)!

Cheers,

Vaughn Stewart