I have the privilege of sharing with you the following report compiled by a rather large customer which compares the results of virtualization consulting engagements completed by EMC and NetApp.
I don’t want to share too many details, as you really should read this report. However, I will share that the results of this report concluded that by deploying storage virtualized by Data ONTAP, this customer would realize the following savings
• 42% less storage
• 58% less space
• 57% less power
I'd like to thank our customer for making their findings public. I can share with you that they hope you will find the data in their report valuable.
I'd say this is another validation that Virtualization IS Changing Everything!
Uh, oh, do I need to change my moniker to VICE?
Few questions or comments regarding this. Firstly, because the customer used NFS, they were able to achieve Thin Provisioning. I am curious as to using an EMC Celerra with NFS would achieve the same results, thus a reduction in power, amount of disks, etc, etc. Doesn't seem like an like for like comparison. This is naturally going to lead one way as there are always multiple ways to provide reduce the amount of disks / datastores required. This doesn't however, highlight I/O requirements, which I am naturally curious about. Would also be interesting to see the comparison using vSphere with Thin Provisioning as well...Just my 2 Cents worth.
Posted by: Kelly | July 24, 2009 at 11:59 PM
Kelly,
The capacity used in NetApp FC and NFS implementations should be about the same. Deduplication an thin provisioning of the LUNS containing thick provisioned VMDKs effectively gives the same utilisation figures.
According to the presentation, thin provisioning accounts for about 18% of the savings. The rest comes from more efficient RAID implementations and deduplication.
Posted by: John Martin | July 25, 2009 at 06:40 PM
John,
I agree, there are savings here, but from Thin Provisioning (thanks to NFS) and from DeDupe. The RAID implementations are different yes, not sure about more efficient though. Agree though that as EMC don't provide DeDupe (yet), with the provided information they need more disk, etc, thus ensuring the power, space, etc increases.
Certainly not arguing here, just highlighting a slight edge towards NFS - which certainly has some massive benefits.
Posted by: Kelly | July 26, 2009 at 04:40 AM
Hi Kelly,
From a technical standpoint, the EMC Celerra makes more sense. However from a practical standpoint - this would not be common comparison if the vendors standard recommendations are followed. For example: 'Consulting engagements completed by EMC and NetApp' suggests just that.
And this follows my own first hand experience where EMC blatantly use FUD against NFS - or anything not FC for that matter.
So it's a fair comparison since that is what many customers will be using from each vendor.
A generalisation I know, but the Celerra option for EMC is treated on the whole like snapshots - not worth even mentioning (???) until a customer brings it up. I.e. the "We have that too" tick box.
Posted by: Paul P | July 26, 2009 at 06:43 AM
Vaughn - in spite of it all, dammit, we will remain friends ;-) Chad/Virtualgeek here (EMCer for those that don't know me).
This is as silly as the original "NetApp useable space" episodes (done buy others), or the "50% guarantee" (done by NetApp). These comparisons always put the "other guy" in the worst possible light.
I bet one is Thick vs. Thin (at the virtual disk layer and/or the storage layer). NetApp talk about it like they're the only ones that do that - news flash (to 3PAR also) - this is now ubiquitous.
LIkewise, NetApp super-fans talk as if every vendor out there only does RAID 10 (at least that's always the comparison) - where of course we all do RAID 5, and many (EMC included) also do RAID 6.
One more? NetApp super-fans talk as if everyone else in the world MUST use BCVs/Clones, when of course the reality is that we can all do snapshots.
They always neglect that many of us can do BOTH (and in the cases where NetApp can't do both - the immediately decry that function as "not needed"), but where the facts don't mesh with the world view of NetApp = best, those things are pushed aside through the mental gymnastics of fanaticism.
None of this isn't to say that NetApp doesn't innovate (no more or less than others), and doesn't lead in some categories (first to market with production storage dedupe - but we're working hard here) - but they also lag in others (and this is not just against EMC, but many others as well): scale-out NAS (soon enough of course), scale out block, enterprise-class SAN, dense storage configurations, spin-down, EFD, non-disruptive movement of storage objects between tiers and the next step in that path - automated tiering.
Look - not saying NetApp shouldn't be proud of their tech and who they are, but that closing a blind eye to what's out there in the market (and listening purely to one's fanboys) is not a healthy state. Sometimes we do it to - but seeing as how we're PERPETUALLY the cast as the bumbling bad guy in these sorts of posts (and in most NetApp customer presentations - where they constantly compare to EMC) - it seems endemic for the NetApp crew.
Stop comparing (incorrectly, even worse) yourself to others - you're in the big leagues, you don't need to do it.
Posted by: Chad Sakac | July 26, 2009 at 01:07 PM
One question as to RAID 6 for EMC systems -- under what scenarios is it recommended? (Not trying to make a straw man as I've been researching this recently and am not finding instances of RAID 6 being recommend for production data.)
Posted by: Andrew Miller | July 26, 2009 at 06:51 PM
Chad,
I'm going to disagree with you here.
From my experience, EMC themselves are IMO the worst offender here.
At least where I'm from, there is still a pervasive 'anti anything not FC'. To the point that in only a recent example, EMC lost a deal, in part, because they insisted that NFS simply wouldn't work - wouldn't do what the customer wanted. This being very funny indeed because the customer had seen for themselves otherwise.
Chad, you are a very respectable blogger...
But I'm afraid your argument, while technically not an issue, is not the point.
Posted by: Paul P | July 27, 2009 at 01:29 AM
@Andrew - thanks for the question. We generally do recommend parity RAID schemes as the default, and increasingly are recommending virtually provisioned parity RAID as the default. The choice of RAID-5 vs. RAID-6 is primarily geared by the size of the RAID Group in EMC parlance, or Aggregates (composed of RAID Groups) in NetApp parlance.
Large RAID groups benefit from double parity schemes to lower exposure during rebuild periods. The downsides to double parity (useable space primarily, but a little more parity math/IO too) are worse with smaller RAID groups (for example, whereas in NetApp land RAID-DP is generally the default, RAID-4 is a good choice with very small aggregates)
Use of double parity with EMC designs is not as critical as it is with NetApp designs since in general we use smaller RAID groups (common in EMC land are 4+1, 6+1 - common in NetApp land seems to be 14+2), and use of things like pro-active hotspares (which start sparing before the drive fails - as the rebuild itself is one of the most stressful time on the drives).
Of course, there are some use cases (a smaller, write-dominated workload) where RAID 10 generally works best (again, not implying anything for NetApp - their different architecture with NVRAM and journalling is different, so I won't hoist the canard of "You don't do RAID 10!")
@ Paul P - I can tell you definitively that we DO sell a lot of Celerra, and a lot of VMware on NFS. I've gone on the record many times myself of course on the topic (I think NFS is great for MANY use cases). But, that's not your point I think - you're implying a more general EMC consensus. I can't speak to the specific situation you refer to, but for your consideration:
1) Whether EMC or NetApp are #1 or #2 in the NAS market is always furiously debated (depends who you ask, and how they measure), but that we're close is not up for debate. Even if I say "we're #2", that still means we're selling a TON of NAS - in the same ballpark. A rationalist can't look at that and say "they're FC heads" - it just doesn't jive with the facts. The customer and the market say otherwise. This is also one of EMC's fastest growing segements of our business.
2) Also, I'm lucky to have a unique perspective on the totality of EMC's VMware customer base. About 30% of our customers are using NFS (some exclusively, most in conjunction with block).
3) For the upcoming SRM beta which adds vSphere support, but perhaps more importantly NFS support - there are 4 times more Celerra NFS customers in that than there are NetApp customers.
Now - let me say this: I think we do have a long way to go, and lots of improvements to make - both with embracing more NAS use cases, and lots of innovations around NAS. That isn't a concession vs. you folks - a respected competitor - rather a concession that there is more do to. None of us are perfect :-)
Thanks guys, and have a great day!
Posted by: Chad Sakac | July 27, 2009 at 06:34 AM
All - thanks for the feedback. Some great points.
On Dedupe and Thin Provisioning (TP):
IMO the dedupe expectation of 30% is ultra conservative as I've never seen a customer with less than ~45% storage reduction. In fairness the EMC config should have included TP VMDKs as well. I will post more details on these two technologies and in order to demonstrate that dedupe delivers savings beyond what is possible with TP.
On RAID 6:
For the past 3 years I have been asking prospects if they have RAID 6 deployed in any type of environment including VMware, but to date my inquiries have been fruitless. I would love to speak to anyone using RAID 6 with VMware.
On Snapshots:
It is well known that EMC snapshots are implemented very differently on all of the product lines and that their usage has a negative impact on the performance of the production data set. This limitation has been validated by 3rd parties many times.
On Comparing Technologies:
I am unaware of a more efficient means by which to help customers to easily quantify the value in competing technologies.
On Fanaticism:
I believe ESX delivers the technology required to virtualize all of the intel based architectures within a data center, and I believe Data ONTAP is on par in terms of storage virtualization. You can think of ONTAP as a storage hypervisor. Am I fanatical? That's a rhetorical question...
Again, thanks to all for the feedback to date. I commit that if you keep reading and sharing your thoughts than I'll keep writing.
Cheers!
Posted by: Vaughn Stewart | July 27, 2009 at 06:52 AM
It looks like Chad and I were posting at the same time and I missed his second post.
Regarding the SRM beta program:
Chad, the VMware SRM beta program has a confidentiality policy in which it clearly states the following:
You may not:
* Discuss this program publicly.
* Discuss this program privately with anyone not registered for this private beta program
(this was copied and pasted from the confidentiality policy)
I'd love to discuss the details of the adapter and our customers' participation, but I cannot.
I believe your claims are incorrect and irresponsible.
Posted by: Vaughn Stewart | July 27, 2009 at 01:55 PM
1/8 = 12.5% overhead - NetApp RAID 4
2/16 = 12.5% overhead - NetApp RAID-DP
1/5 = 20% overhead - EMC (RAID5 4+1)
1/7 = 14% overhead - EMC (RAID5 6+1)
2/10 = 20% overhead - EMC RAID6 (8+2)
2/12 = 16.5% overhead - EMC RAID 6 (10+2)
The last 2 configs are EMCs recommendations and constitute what EMC calls a "medium" RAID Group and performs "optimal".
You see, in a typical RAID 6 config such EMC's, the second parity drive not only adds overhead but more importantly, the additional parity computation, can have a negative effect on I/O. Particularly for a workload with Random I/O characteristics.
Now, anyone who has run perfstat on netapp box they will notice something quite bizzarre...and that is that the most I/O *underutilized*, drives are always the parity drives...and that's why NetApp recommends RAID-DP and others shy away from RAID 6.
I recently saw a NAR output off a clarrion were the customer had 100 Raid groups some of which were in 4+1 RAID 5 configs and others in 7+1. Right off the bat the customer had 100 parity drives and close to 590 drives total (excluding host spares).
A similar config with NetApp (42 shelves x 14 disks) would have yield 74 Parity Drives WITH RAID-DP - total.
So yes, EMC, and not only, is using Smaller RAID Groups, they just use a ton of them.
Here's something to ponder...RAID 5 - 6+1 Raid Group...Lets use 300GB FC drives...
((300,000,000,000 * 6 * 8)/ 10^15) = 1.44% double disk failure.
How 'bout SATA...500-750GB is typical...RAID5 = 4+1 (10^14 unrecoverable bit error)
((750,000,000,000 * 4 * 8)/10^14) = 24% risk of double disk failure
These number are large. Even 1% is large, if you can avoid it without performance implications.
Proactive sparing, scrubing, checksums are techniques widely deployed by most disk arrays today but that does not eleviate the need for a capable RAID 6 solution given some of these techniques (scrubing) cycle thru the entire system and the process can take DAYS to complete. So while the scan occurs on, say, Device Bus 1, Enclosure 3, Disk 3 ...Enclosure 10, Disk 10 might exhibit an issue rendering ANY type of scrubbing solution absolutely useless to address this particular problem.
Posted by: Nick Triantos | July 27, 2009 at 03:27 PM
Wow
Just a few days ago, I upset more than a few NetApp fanbois for claiming that NetApp's conduct lends itself to unprofessional and cheap marketing tricks -- gaming benchmarks, unfair comparisons, worthless guarantees, etc.
Now we have this latest effort as yet another prime example.
You're misleading people once again, folks -- not in one or two ways, but in a whole bunch of ways.
And, despite this, I'm sure that every NetApp rep will have these "customer" slides in front of as many prospects as possible -- facts be damned!
We at EMC do respect your products -- but we don't respect your conduct in the marketplace.
And here's yet another example to join the others.
-- Chuck
Posted by: Chuck Hollis | July 27, 2009 at 03:34 PM
It never sizes to amaze me in this industry how a vendor who can't win a specific battle, will "cry foul" and "play the victim" in an attempt to gain sympathy points...
Instead of Mr. Hollis providing a counterpoint(s) based on this customer's published findings he has chosen instead to play...the "victim".
Posted by: Stephan Sebuit | July 27, 2009 at 07:37 PM
Chuck and Chad,
I concede, the two configurations the customer looked at were different but this point is irrelevant.
The customer received the best offerings available that were inline with their budget. These quotes were designed along the best practice recommendations from each vendor.
The customer choose the technology which allowed them to virtualize more with VMware.
The customer was the winner!
That's all that matters, not if arrays from NetApp and EMC are configured identically.
Posted by: Vaughn Stewart | July 28, 2009 at 07:16 AM
Hey Vaughn, a few weeks ago you posted on an epiphany you had and remarked in it:
The 18 point gain is a growth in market share of 300%. By contrast EMC's growth from 42% to 48% is a 6 point gain or a growth of 14%. In addition, NetApp is in 3rd place by 1% behind IBM which OEMs the N-Series arrays from NetApp. I can't wait for the findings reported in the 2010 reports.
I guess we see the real problem EMC is having, and this current post shows why this is happening - I'm guessing.
It is interesting when you back up, and take in the big picture, what we see - no?
Posted by: Mike Shea | July 28, 2009 at 07:52 AM
The 2 surveys asked different questions. One asked to name one vendor. The second accepted multiple responses. The two can not be compared. The 2nd survey adds up to way over 100%. Comparing 2 surveys with different parameters is nothing more than an intentional mis-representation of the facts. (Nothing new for you guys) (NetApp is in Test Dev / tier 3 / File sharing in many large Enterprise Envionments that employ a dual vendor strategy, EMC is Tier 1 and 2. Oracle Corp. comes to mind with this model) I think this is the best explanation for the difference in the numbers. I think you may want to update your blog regarding this "Epiphany" It's purpose was nothing more than to deceive people. You are obviously intelligent Vaughn and missing that fact is something that would cause you to fail an 8th grade math test.
Posted by: RDANIE01 | July 28, 2009 at 10:33 AM
RDANIE01,
Iwould like to ask you to read the original 'epiphany' blog post.
The data being represented is from an EMC presentation where they use these values to communicate their dominance in the market.
I believe your comments should be directed at EMC and not NetApp as we are merely reusing the data they train their teams to distribute and promote.
Posted by: Vaughn Stewart | July 28, 2009 at 10:50 AM
My problem is with the 300% growth conclusion by NetApp. That is a factually false statement.... no? If the customer came to this conclusion on his own then I assume he is not building budgets or TCO's because the CFO will have him fired in short order. The Market dominance is a true statement given that in both surveys they are #1....By a wide margin. EMC didn't take these 2 surveys and claim to growth, rather they took two surveys, and in some cases 3 surveys that all show the same thing.....EMC being the clear market share leader. Using multiple surveys validate their claim and smooth out statistical inaccuracies of any one particular survey. I think this type of mis-information is what Chuck is describing as NetApp's "Market Tactics".
Posted by: RDANIE01 | July 28, 2009 at 11:22 AM
Wow Chuck,
You know, one of the stages of denial is projection. It was just what, 3 months ago, when EMC released this? http://www.emc.com/collateral/hardware/technical-documentation/300-006-724.pdf
In it, EMC configurs a 5,146GB storage pool, from which they carve a 300GB Celerra file system and 300GB iSCSI LUN for use in VMware NFS performance testing. Massively short stroking for a performance test (only 5.8% of the pool is used); what would you call that? Cheap marketing tricks? Gaming benchmarks? Unfair comparisons? I'd say you're in denial. You're projecting EMC's tactics on its competitors.
John
Posted by: John F. | July 28, 2009 at 12:15 PM
@John F - that's a technote. It basically reports the results of a test, and makes some conclusions.
If we were to look at that and say: "Ergo, we are better than X" that (I think) would over the line. There's nothing even remotely like that in there, is there? The "find flaws in vendor whitepapers for the purposes of competitive bludgeoning" is so... i dunno, silly? Now, if we've made an error (and we have in the past, and I'm sure we will in the future), finding an error and correcting us - that's not only in bounds, it's what we should all try to do.
@Vaughn - you're right about the beta note, I'll apologize to the VMware SRM team. First rule of fight club = don't talk about fight club. I also shouldn't have made a conclusion based on the number of participants (particularly as it varies during a program, as i'm sure is the case here). I got frustrated, and crossed a line (as I personally feel is being done here in general). My point I'll re-iterate, NFS is very important to EMC, as of course it is to NetApp. The thing that got me frustrated was one of the NetApp folks tarring us with the "FC-bigots" brush. That's simply not who EMC is. I've consistently been a strong supporter of NFS as a storage model for VMware - though not in all cases (as has also been your position on our joint post).
@ Nick - The N number of R5 parity disks in that configuration was the how that customer was configured. We of course support very large RAID groups, and of course support R6. Every customer is different. My point on scrubbing and hot-sparing and proactive sparing was that every vendor does this differently. I have access to my NetApp filers, and I'm sure you guys have access to CLARiiONs, Symms and Celerras (the Celerra is openly available for download for playing with, so anyone can).
Like it or not, our storage business ain't the server business. The platforms (while all using similar commodity elements) are all SO distinct and different from one another - direct comparison (between NetApp, EMC, or anyone for that matter) is almost impossible. Overall comparisons are difficult, but in the end customers make choices. Our Q2 results speak well - customers continue to choose EMC (since we're all so sensitive, I want to be clear - I'm not claiming it's at the expense of anyone in particular).
Again, to me, it's an example of the thing that made me comment here in the first place. I'm not trying to cry foul, just suggesting that the tactics of constant negative selling.... well it's almost like the disease that eventually doomed the Republican party. Eventually, it just reflects poorly on the source.
Do you guys REALLY want me to post swap out stories in the other direction? Man, I don't. What a waste of time. Both NetApp and EMC sometimes fail to serve our customers adequately, sometimes due to technology, and sometimes due to other factors. In spite of the fact that I do track these, and have many examples, and just like you have a "bad NAR" case, I have a "bad perfstat" set of cases. Personally, I don't think it helps to drag anyone (NetApp or otherwise) through the mud. I try to maintain that - not perfectly, but I try.
@ Vaughn - re postioning of the studies re: what customers were doing - you know very well (and you know why I know you know) how we train our field to talk about that: "Here are three studies, totally different customers, totally different vendors, totally different sample sizes, three different years. You can't draw any conclusion from that except for a lot of customers choose EMC for VMware environments - and let us tell you why, based on our solutions" (for some reason I don't know you picked 2 of the 3 studies we've used). I can't defend someone who positions it as "Marketshare" or as "trending" because they are not, and we've repeatedly asked our field NOT to use them that way.
Doubt it? Read this public post from ONE YEAR AGO!!!
http://virtualgeek.typepad.com/virtual_geek/2008/07/whats-in-a-stud.html
We're using it because certain folks have trained their folks to say things like "we're is more integrated with VMware than even the parent company!", or "we're the NFS reference platform", or "VMware runs their IT on ______". It's not true, and it's vendor sillyness, but people LOVE the idea (just because they also love to read about what Michael Jackson ate for dinner the day he died) that there's some sort of VMware/EMC feud, or that the "plodding East Coast company is riding on the coattails of their West Coast subsidiary" or whatever. It's simply not accurate.
Here's hoping we can all be the change we want to see in the world.
Nick, I'm extending the same offer to you as I did to Vaughn. At VMworld, for a bit at least, let's put down the weapons, learn a little about each other's technologies. I'll do it - hope you guys are willing to do the same. Afterwards, I've already committed to a round, and you are welcome to join.
Posted by: Chad Sakac | July 28, 2009 at 05:19 PM
"The thing that got me frustrated was one of the NetApp folks tarring us with the "FC-bigots" brush."
Hi Chad. Actually I don't work for NetApp - but I do like their gear.
Sorry you don't like the generalisation - while I don't oppose it (I've not come across an EMC customer using NFS for VMware, yet. Not to deny that it doesn't happen somewhere in the world) - in this case it was commentary from a potential EMC storage customer, sorry to repeat myself: Was simply amazed that in spite of their interest in an NFS storage solution for VMware, that EMC insisted from the get go "why wouldn't you use FC", not only dismissing the customers own interest but also spending time explaining why NFS would not work so well.
This potential customer obviously did not read your blog (but that ain't their fault is it?)
So please forgive me Chad - I'm left wondering exactly when EMC WOULD recommend NFS for VMware storage....?
Posted by: Paul P | July 29, 2009 at 02:54 AM
It is our general recommendation, Paul.
In all our training for our field, our recommendation is simple. Almost every customer can benefit from leveraging both NAS and block protocols together in VMware use cases. There are distinct pros and cons, but that doesn't translate into a forced choice for a customer - they clearly have choices (both from NetApp and EMC, and others also) where they can leverage both together.
My apologies also for assuming you were a NetApp employee (many folks in the comment thread are).
And for what it's worth (I'm just one person after all) - I'm aware of hundreds of EMC customers using NFS storage - either exclusively (where they don't hit the use cases where NFS is less than ideal) or in conjunction with block (sometimes with EMC block storage and sometimes non-EMC).
Would love to know who the EMC account team was, and I'll happily have a chat with them.
Posted by: Chad Sakac | July 29, 2009 at 05:04 AM
@cChad,
I'm glad you've cleared up the recommendations from VMware, however I think you missed the point in my rather sarcastic last post.
Chuck sees this as the result of some nefarious activity on the part of NetApp to misrepresent something. In fact, it's the result of a poorly designed EMC test puslished in an EMC document that recommends on/off configurations of VMware on Celerra NFS that resulted, apparently in this case, in EMC recommending the configuration that they did.
What I did for Chuck was point to the smoking gun that resulted in the competitive situation playing out the way it did. If there's a lesson in this for EMC, it would design tests and do a bit more fact checking than you see in supermarket tabloid articles before publishing recommendations. Sure, everyone makes mistakes. I just don't understand why Chuck insists on creating conspiracy theories that place the blame on NetApp for something that is clearly an EMC mistake.
John
Posted by: John F. | July 29, 2009 at 10:02 AM
Wait... is Virtual Provisioning not the same as thin provisioning? Why would you compare the space requirements of a thinly provisioned system versus a thick provisioned one. Also, don't some of us still use RAID-10 for a reason? Some of my LUNs are RAID-10 provisioned, and it's not because I want to waste physical space, output more heat and pay more to my datacenter provider each month. With the exception of deduplication of primary storage, what was the advantage for NetApp - my EMC stuff does all of that. And while I would like to explore deduplication for some of my tier-2 data storage needs, I don't think it's the be-all, do-all, end-all solution for all of my storage. We run it now on an older FAS-3020 and it chokes our box from a CPU utilization point. No doubt higher end and more modern NetApps do a better job... but there's clearly a performance hit that maybe outweighs things like consolidating cabinets.
What is shown above is an extreme case. Yeah, in a perfect world I'd do 15:1 or more on my VMware boxes, and go with super dense drives (1tb SATA is really chock full of space) and thin provision everything and dedupe everything and use 1gbps iSCSI since it's cheaper than my FC links and way cheaper than those 10gbps links, and I'd use DPM to power down EVERY box in my cluster not needed so I could over provisioning my hosts and run everything pegged at 100% CPU, and I'd even take my bicycle to the data center next time there was an issue, sure I'd get there slower in an emergency, but our carbon footprint would drop. Wait... no I wouldn't!
Posted by: Scott Lewis | August 01, 2009 at 07:28 AM
Scott,
The answer to your question above is no. EMC Virtual Provisioning is not the same as NetApp Thin Provisioning. EMC makes this point quite clear in their best practices guides. Example:
"Conceptually, the thinly provisioned storage pool is a file system overlaid onto a traditional RAID group organization of hard disks. This file system imparts an overhead on thinly provisioned LUN’s performance and capacity utilization. In addition, availability must be considered at the storage pool-level with thin provisioning. (Availability is at the RAID group level with traditional LUNs.) Workloads requiring a higher-level of performance, availability, and capacity utilization should continue to use traditional LUNs."
-- EMC CLARiiON Performance and Availability: Release 28.5 Firmware Update Applied Best Practices
Not only is EMC capitulating to the need for a "file system" layer to provide such a function (ironic given all the FUD they like to throw around about WAFL) but - and this is striking - they admit that you get BETTER availability, BETTER performance, and BETTER storage efficiency by NOT using virtual provisioning.
NetApp recommends Thin Provisioning for better storage efficiency without sacrificing performance or availability. Huge difference. NetApp is a huge fan of TP while EMC is not. Example:
"I think thin provisioning is not-a-good-thing at a philisophical level. It has a role, but I'd recommend using it very carefully, if at all."
-- Chuck Hollis, EMC VP of Global Marketing
I believe Chuck is sincere in his belief not to mislead customers. I think that's exactly part of the NetApp culture as well. From that standpoint you would have to take EMC documentation and Chuck at their word: don't use Virtual Provisioning if you care about availability, performance and storage efficiency.
As far as RAID-DP vs. RAID-1 is concerned, you can look at any number of 3rd party benchmarks both solicited and unsolicited by Netapp. NetApp provides as good (or better) performance compared to RAID-1 with statistically higher data availability using RAID-DP. It's not marketing spin. It's just math.
Now, if a customer wants a RAID-1 type of config, NetApp can provide that too. We do have customers running this config as read performance can improve slightly and the customer can sustain 5 concurrent disk failures within a RAID group and still stay up and running with their data intact.
Can EMC and NetApp give the same short answer on an RFP? Sure. But, the devil is in the details so, no, I would not agree that Virtual Provisioning is the same as Thin Provisioning. There's a big qualitative difference there. Unfortunately for EMC, they are slowly starting to realize that it's difficult to bolt-on and shim-in these "file system" services after market to all of their sundry "purpose built" storage systems.
Posted by: Mike Riley | August 01, 2009 at 02:30 PM