Fun with Vendor FUD – Episode #1

UPDATE Feb 12th, 2010: 10:52pm ET:

I really do want to reiterate, I posted this in a light spirit, some of the claims struck me as funny ("The only SAN with vSphere vStorage Fault Tolerance support included at no extra charge"), and some so demonstrably incorrect ("The only SAN platform which supports SRM for both automated failover and FAILBACK") that I laughed, and my team laughed.

So, why not spread the laughter?  Don’t read into any more than a fun thing to make us laugh.   And yes, if you get very literal, it’s not technically FUD if it’s not directly negative on another vendor, but hey, then the alliteration in the title of the post doesn’t work :-)

UPDATE Feb 17th, 2010: 1:16am ET

Thanks for all who responded!   I’ve closed the survey, and gathered 209 responses from around the world.  The answer to the question “who said this?” is below.

There’s so much fud that gets slung around these days (not implying EMC is immune!!!) that rather than get frustrated by it, my team tries to have a bit of fun with it.   I think the game is fun enough to share out there :-)

Ok, this first variant we call “name that vendor”.   A question comes in from a customer, which is clearly a “list of ‘unique's’” that was sent by a vendor.  By “unique’s” I mean a list of pithy points where a vendor makes a grandiose claim of uniqueness on a given feature in the hopes the customer makes that one thing on that list a buying requirement.

What we like to do is strip out the customer name, and the vendor source, and play “name that vendor”.  It’s always fun because it’s usually from sales, and they so mangle what the product actually does do that’s cool, or make just flat out incorrect or silly claims.    Since VMware is one of the most popular customer use cases of infrastructure, we get to see some fun ones…

Here’s the first Virtual Geek episode of “Fun with Vendor FUD – Name that Vendor”! (I’ll answer the source after a week of polling)

• The only SAN with vSphere vStorage Fault Tolerance support included at no extra charge
• Full Application aware API integration with VCB at no extra charge
• Clustered SAN architecture to match your ESX server cluster
• Fully integrated MS VSS apps with API hooks
• Full Synchronous replication software included at no extra charge
• The only SAN platform which supports SRM for both automated failover and FAILBACK

Survey has been closed – don’t worry – there will be more episodes of “Fun with vendor FUD” – survey result (209 votes) – results below (including geographic dispersion of votes)

There’s all sorts of vendor craziness out there.  

My advice?   Ignore claims of “uniqueness” (heck, even from EMC in general, or me specifically).   Look at the product on it’s own merits.   What problems could this help you solve today?  Does it have value that is greater than it’s cost.   Does it align with where you see IT going?  

“uniqueness” in any industry is hyper-transient.   “Different”, well – every product has architectural elements that mean that it can do things differently – and that’s legit, and customers should consider them.   My other advice?  Have some Fun with Vendor FUD :-)

If you think this is funny like I do, I’ll keep posting more episodes…

…So – who said it?  

First of all – don’t take this as an indictment of any vendor – all of us do these shenanigans periodically.   And also, remember that this truly is in the spirit of fun.  I will try to be even handed, including “Fun with Vendor FUD” episodes that poke fun at EMC.

… so – in this case, the source was from an HP/Lefthand.   The giveaways (to me) were:

• “The only SAN with vSphere vStorage Fault Tolerance support included at no extra charge”.  I think this is a rep confused about the fact that you can run the Lefthand VSA, and apply FT against it (though of course, they software RAID and have a distributed volume manager, so you don’t use it that way – you run multiple VSA nodes?).   vSphere Fault Tolerance works on all storage platforms on the VMware HCL.
• “Clustered SAN architecture to match your ESX server cluster”.  This one is a funny one to me too – while there’s an argument to be made about "scale out of x86 servers with DAS with software turned into a scaled out storage array” (note that interestingly a V-Max is literally a set of scale-out x86 server with storage attached – just with the idea of linking all the servers into a common memory space so cache is global/any IO served by any port/brain via the Virtual Matrix) is a good design, there’s nothing implicit that makes that better for vSphere use cases, more or less than any other use case.  Remember I came from a company (Allocity) that did a similar idea (though at BEST we were a much, much less mature implementation than Lefthand), so I get the value, and the downsides of the approach.  

This one is flat out wrong.

• “The only SAN platform which supports SRM for both automated failover and FAILBACK”.   This is wrong, EMC offers this on 3 of our 4 SRAs, and I believe that EqualLogic and Compellent offer something similar.

Look, I’m not trying to stir the pot here.    Honestly, every vendor does this, and every sales rep is always looking for “silver bullets” to put out there.  I’ll continue the series so long as it’s in good fun, and people don’t get too bent.  I’ll try, as always to be as even handed as possible.     Was interesting to see the votes – NetApp and EMC got lots of votes (though the statements too me at least pointed to EqualLogic or Lefthand) – suggests perhaps we both have a reputation for a lot of FUD?

Comments

[Alex McDonald]

FUD? I don't think so!

There are only two specific competitive claims in that list; they start "The only SAN..."

If those claims are wrong, then it's either (a) a lie or (b) a mistake. And easily checked, verified and then proven or disproven. And some of these are no more than word salad marketing claims; like "Clustered SAN architecture to match your ESX server cluster"

This is FUD. At its stinky and rotten finest; " But RAID DP gives you the best of RAID 1 and RAID 5 - for the first few hours."

Who said that? I'll save you the bother; HP.

[BradC]

I'm still trying to figure out what "vSphere vStorage Fault Tolerance" is :)

[marc farley]

I think its very funny, but probably for a different reason. Exposing the practices of sales people would bring new lows to the blogosphere. Perhaps you think EMC does less of this?

Uniqueness happens and it matters. Netapp has had it for close to 20 years with WAFL. There are strengths and weaknesses to discuss with WAFL, but I'd still call it it unique. I've long admired their ability to leverage that technology for so long. SVC from IBM - definitely unique! Chunklets from 3PAR (my employer) just as unique.

Product breadth helps EMC align with a lot of what is going on in the industry - and I would say they are uniquely positioned to do that. Nobody else in storage can touch EMC in technology breadth - not Netapp, not IBM, not HP and certainly not 3PAR. But that doesn't necessarily help each individual EMC product solve customer problems better. On the contrary, I suspect it creates some amount of burden.

[Mcfm_pt]

I don't know all the products from all that vendors, but are you talking about IBM? They are saying lot's of "...at no extra charge" things about XIV product.

[Chad Sakac]

Guys- I really do want to reiterate, I posted this in a light spirit, some of the claims struck me as funny ("The only SAN with vSphere vStorage Fault Tolerance support included at no extra charge"), and some so demonstrably incorrect ("The only SAN platform which supports SRM for both automated failover and FAILBACK") that I laughed, and my team laughed.

So, why not spread the laughter? The practices of Sales people will bring humor, not new lows :-)

BTW Marc, I put EMC on the list for two reasons: 1) EMC absolutely DOES this as much as the next guy, and 2) because heck, one of my team members' votes was "this sounds like a really bad block-only CX sales dude"

Also, maybe I can be more clear on my comment on uniqueness too - architectural models in storage are in fact remarkably unique. Remarkable to me at least that a common function (presenting a file, or presenting a block device, or an object device) have such WILDLY different internal implementations. Think conversely about how little internal architectural variation there is in servers, or even networking (or at least that's the way it seems to me).

That's perhaps why DIRECT storage comparisons are relatively hard.

The point I was trying to make is that major functional (not implementation-level, but functional level) capability over time reaches across the market. There will be variations in the implementations (with the core architectures lending themselves to pro/con implementations.

BTW - there is a long list of EMC uniques, beyond just breadth, but I'd agree with your point that we @ EMC can do more to integrate the value across the breadth of portfolio. For the last 2 years, it's been a heavy effort, and 2010 is the first year where those will start paying dividends in a large way for our customers.

I'll drop a hint, the current votes count winners are NOT the source :-)

[Calvin Zito]

Alex - I'll be even more specific about who said "But RAID DP gives you the best of RAID 1 and RAID 5 - for the first few hours." It was me.

As you know, we've tried to reproduce your ESRP results and when we let the test run for more than a few hours, the performance degraded (http://bit.ly/lxNOV). We got no where working with NetApp to figure out why - the last answer was "Maybe your FAS has a problem and needs to be fixed". I'm not trying to open up that debate again because I know it will get no where.

But what got my attention was the Exchange benchmark that VMware ran - an intense load over the course of more than 8 hours. And low and behold, what did those results show? When the test started, IOPS for the FAS6020 were 9000; at the end of the test, the IOPS were between 3 - 3.5K. See http://bit.ly/axD9CN, page 8. Since the only response I've seen from NetApp on this is name calling, I can only assume it isn't FUD.

Have a good weekend.

[Alex McDonald]

Hey Calvin, how about posting my reply that I made on your blog? Or are you having some local difficulties with the EVA that stores the posts, and you've done some kind of rollback?

http://www.communities.hp.com/online/blogs/datastorage/archive/2010/02/11/super-bowl-ads-and-capacity-guarantees.aspx

[MauroAyala]

WAuuu, nobody want to laugh here? to me it is so funny and is a good training trying to figure out the correct answer, Clustered architecture sound like Lefthand, EqualLogic and Netapp (in my opinion).....At no extra cost sound as Lefthand, EqualLogic and IBM .....automatic failover and failback EMC. :)
I am Technical Consultant and want to play the game don't you? my be the answer is EMC who knows !
rianse un poco :)

[Dikrek]

@ Calvin:

Any array degrades as it's filled up, and Kostadis Roussos explaid the NetApp aspect of this in detail in http://bit.ly/cnO2.

How about showing us the same test with a similar EVA doing RAID-6 (to get the same protection) and like 100 snaps active (since NetApp customers would be doing that as a matter of course).

You see, if you have nothing to compare this to, all you have is a graph for a single product, and your assertion, while seemingly correct, means nothing unless compared to something else.

D

[Dikrek]

Back to the original subject of the post:

I understand why some of the smaller (some would say irrelevant) vendors make these assertions: ANY publicity is good publicity!

Check here for some craziness from a vendor that hasn't managed to secure decent marketshare in a while: http://bit.ly/bJSvRr and here http://bit.ly/aSbEED

D

[John]

Why does everyone assume Netapp ? maybe were just so used to these type of outlandish claims.

[TimC]

Calvin, I really hope your job isn't technical in nature. Please, review page 8 again, and don't just review the top graph of IOPS. Take a look at the second graph. Notice those response times? If the storage array was the limiting factor in the IOPS being served, the response times would be spiking. The host side requests would continue at X IOPS, and the array, unable to keep up, would have a latency spike.

CPU's aren't pegged. Drive's aren't lagging... please, enlighten me as to what part of the storage array is limiting those IOPS. If it was "WAFL fragmentation", that disk latency would be absolutely through the roof.

http://bit.ly/YzVlR