National Storage Rip-Off Day, Part Deux

by Administrator on January 13, 2008

So many responses to my earlier post on this, I had to open another thread.  Basically, I got hammered by a few guys for comparing EMC/IBM/HP/HDS wares aimed at SMBs to less expensive products from Zetera, Promise Technologies, Hifn, etc.  Apples and oranges they said.

Complaints came down to several things.  Some said these much more expensive collections of spinning rust from brand names have a lot more testing behind them.  They come with one stop shop hardware and software builds.  They have big cadres of support folk behind them and even phone home if a problem arises.  And they are certified to work with various OS’s and application software, a pricey but useful discriminator.

I could do (and have done) a big rant on a lot of the so-called certification programs out there that are complete and utter BS.  Pity that some believe that they have value, forcing the small guys to have to “pay to play” — to spend precious resources they could be using to build market and sales channels to buy access to proprietary APIs.  It is a huge rip-off unto itself.

TCO wise, I have never seen demonstrated the purported value in buying products that are “certified” with third party wares versus those that aren’t.  On the one hand, VMware says that it can use any LUN exposed by any storage device.  On the other hand, they want storage vendors to pay ($30K per SKU) to be “certified.”  Doesn’t this strike anyone as odd?

There is no rocket science in storage.  Standards exist at the device level, the array level, and to some degree at the interconnect level.  Where these standards don’t work, it is because the vendors have developed the standards to ensure that they don’t work. 

Management remains the huge gap.  Vendors don’t want their products to be managed in common because it gives the impression of commoditization and enables folks to unplug vendor A and plug in vendor B at the drop of the hat.

As for after sale service and support, we are looking at huge fees for something that could be purchased from third party service providers, local integrators, or local aftermarket service organizations just as readily as from the equipment vendor itself. 

I feel very uneasy about the fact that phone home functionality seems to be more frequently used today and that guys show up at the door with replacement parts because the equipment summoned them.  Why is there always an uptick in this activity when new equipment sales slow down?  There might not be any correlation, of course, but if users start to believe there is one, what do you think will happen to the idea of phone home being a value add?

Finally, we have all-in-one on array software stacks that some readers here regard as the real discriminators between the big brands and the little guys.  I hear this one a lot:  “I don’t want to cobble together my own set of storage management software.  I want it all to be delivered on the array.”  Alternatively, this comment takes the form of “I don’t have a huge IT staff, so I want my storage to be drool proof.”  I agree that there is merit to this view for small shops.  I also have to say that acceptance of lock-in proprietary functional sets at a huge mark-up is not going to serve the department as it grows, as applications and corresponding storage infrastructure diversifies, and as you need to make more capacity remain manageable by fewer folks.  Fielding a bunch of proprietary platforms, each with its own proprietary management interface, becomes problematic and costly over time.

Architecturally, the storage device may not even be the right place to host some of the value add functionality.

Finally, misapprehension abounded about RAID levels and mirroring/WAN mirroring capabilities (or lack of support thereof) by the equipment alternatives I cited in the earlier blog.  The commenters didn’t know very much about these products or their support for preferred functionality and simply assumed that they didn’t support them.  Some investigation is in order on their part and I invite Zetera, Promise, etc. to use this blog to post some responses to the intelligent questions that were being asked.  Since it would cost these vendors nothing more than a little time, it’s the cheapest form of customer education I can think of.

Anyway, thanks to everyone for writing and voicing their views.  Keep it up!

 

{ 5 comments… read them below or add one }

jhutchins January 14, 2008 at 12:10 pm

Jon:

You said: “TCO wise, I have never seen demonstrated the purported value in buying products that are “certified” with third party wares versus those that aren’t.”

The value, from the end user perspective, is that VMware support won’t talk to you if they discover that you are using non-certified hardware. I am in no way validating this absurd policy, but the reality today is that I don’t have a choice as an end-user IF I want VMware support. Yes, I could use third-party support, but if anything blows up in my environment and my bosses discovered that I went around the cert process, I’m toast. I totally subscribe to your philosophy of this, but I also have a job to keep.

What we need is major vendors (like HDS, HP, IBM, etc.) to step up and call fowl on the EMC/VMware cert process.

On propietary sytems: I heard an interesting thing recently. A big bank was talking to another big bank about VTL. Both banks were EMC shops and one had purchased their CDL. When they discovered that the CDL (which will ONLY WORK with EMC disk) was really just a re-branded FalconStor VTL, they pulled out the CDL, put in a FalconStor VTL and basically set up an environment where HDS and EMC duke it out now for the disk business. They estimate that they have saved about 65%. The second bank tried to do this until EMC put the brakes on it and said they would not certify their disk to work with FalconStor. Now, the second bank is faced with doing a full rip ‘n replace of the EMC but they aren’t sure if management will have the balls to do it.

The proprietary BS can be broken if enough companies step up and stand up to companies like EMC.

jaded January 14, 2008 at 6:28 pm

I am in agreement, if Vendors really wanted to be friendly towards end-users, they would not use certification programs as a joint marketing scheme and cash cow. At the end of the day it really hurts the users who are looking for choice. And the innovative, non-venture backed companies with more limited resources. I’d like to see a day where management functions like Aperi (and SMI-S) could be widely supported, but I think the trends are there for a new storage paradigm.

If you have noticed, many of the of the popular, new and recently acquired companies are using x86 architecture to build scalable storage infrastructure. This commoditizes the disk, and puts value in the management software. Of course there are open source projects like Open-iSCSI looking to displace some of the other iSCSI targets and Cleversafe to provide a storage grid, and of course open-source implementation experts like SpikeSource, building your own storage appliance isn’t far off. An instant storage appliance, using virtualization to distance the management from the hardware would provide users with the management they need and the flexibility to choose their own support structure.

Tlud January 17, 2008 at 5:41 pm

Response from Zetera (and this should generate some interesting response) … Every Zetera system ever shipped, currently over 200 thousand, includes a storage management system, the ability to create RAID 1 mirrors, and the ability to enlarge any volume at any time through our spanning process. In addition, all of our business class systems include the ability to create RAID 0 striped volumes and the ability to combine striping, mirroring and spanning for a very high level of flexibility.
We have chosen to not support parity based volumes because of the high degree of risk that they pose. For example, a Seagate Barracuda ES.2 1TB disk drive has a non-recoverable error rate of 1 sector per 10E15 bits read. To rebuild a 4 disk stripe from 3 surviving disks, the system must read 3 disks x 10E12 bytes/disk x 8 bits/byte, or 2.4x10E13 bits of data. This means that 2.4% of the rebuilds will encounter a secondary failure resulting in nonrecoverable data loss. We believe that parity was a good solution for smaller drives, but is unsafe when used with today’s higher capacity drives.

Administrator January 17, 2008 at 5:55 pm

Thanks for chiming in here, Tom. I thought no one from Zetera was reading this blog :-).

jhutchins January 18, 2008 at 11:39 am

Tlud:

****
To rebuild a 4 disk stripe from 3 surviving disks, the system must read 3 disks x 10E12 bytes/disk x 8 bits/byte, or 2.4×10E13 bits of data. This means that 2.4% of the rebuilds will encounter a secondary failure resulting in nonrecoverable data loss.
****

True for RAID 5, NOT TRUE for RAID 6!

Previous post:

Next post: