StorageTexan's Blog

The time for Storage Blades.

 I’m writing this blog from the El Dorado Resort in the Riviera Maya Mexico while sipping some sort of Frozen Blue Hawaiian cocktail, so my life at the moment doesn’t suck.  When most people are catching up on the latest Twilight novel, or Andre Agassi’s biography, I’m flipping through whitepapers and tech docs I’ve been printing out for the last few weeks.  For those of you that do a great job of “unhooking” from work during vacations, I applaud you. 

I was reading through a few things, and it’s becoming clear to me that the industry is at a crossroads.  I liken this to the same situation the server vendors went through just 6 or 7 years ago with the deployment of server blade technology.  I recall talking with a large independent school district in West Texas back then about connecting the new IBM BladeCenter with our Magnitude 3D solution.  The BladeCenter was a new thing, and this customer was searching for an ability to reduce rack space as well as cost predictability when deploying future applications.  Since then, blade servers have proliferated to the point that they’re typically the server of choice for deploying server hardware.  Virtual servers have taken this to another level.  It’s not uncommon for companies to fully understand the number of virtual machines a certain model Blade can support and buy different types of blades based on those requirements.

I think the storage industry is at this pivot point as well.  We can accurately tell you within a few GB’s the capacity predictability of a solution we quote, so why is it that we have a difficult time being that predictable with performance?  We still get speechless when customers ask what sort of performance predictability can I expect today, in 1 year, in 3 years and in 5 years.  This is going to become MORE important as virtual desktops and other IOP-hogging applications become more mainstream.  I’ve stopped being surprised when talking with channel partners as they recite problems when they deployed a certain storage vendor’s solution and they find out 3 months later it was undersized. Een though according to the vendor’s tools, it was sized according to the vendor’s best practice, and the customer still runs into performance issues as they ramp up the use of the solution.  What’s even more surprising is the storage vendor’s near-giddy response when telling them how they can solve this issue.  SSD anyone?  Can you spare a few tens of thousands of dollars after you’ve already blown your budget on the acquisition to purchase 2 SSD drives? Just to solve a performance issue?

Maybe it’s time for the storage industry to look at Storage Blades as a means to meet this performance/capacity predictability paradigm.  Imagine looking at storage as a building block for an overall storage strategy that allows you to be predictable in both capacity and performance so your sizing of application deployment becomes easier and more cost effective. Imagine knowing the dollars per IOP per Rack Unit on the life of the storage blade?  In other words, you can expect X IOPS per RACK Unit for the 5 year life expectancy of the storage solution.  Imagine being able to do this knowing that the predictability is based on running the storage solutions at 97% utilized  This begins to put you on par with the capacity planning you’ve grown accustomed to knowing.

I believe that Storage Blade architecture has the ability to bring the storage industry to a crossroads.  Do we continue piling on features and functions, already found in typical applications and OS’s, and live with horrible utilization rates and the inability to predict basic performance results?  Or should we stop this controller feature crawl and hit the big issues at hand: cost, capacity and performance predictability?  Storage Blades, like their server brethren, can possibly be the means to this end.  If nothing more, for certain applications, Storage Blades just make perfect sense. 

Oh well – I just saw someone order a Frozen Mojito that looks really good.  Back to the grind that is my vacation 🙂

@StorageTexan

The COOL things Xiotech is doing with RESTful API !!

(Video Demo at the bottom)

So, if you remember a few weeks ago I blogged about Commvault’s launching of their RESTful interface connector.   Well, one of the cool things to come out of this spring’s Storage Networking World in Orlando is Xiotech’s release of our CorteX Control Path RESTful API.  As far as I know, Xiotech is the only Storage Vendor pushing this sort of announcement out.  For cloud providers, this is the ultimate in flexibility around monitoring and managing storage blades.   Just as an example, in the Xiotech booth we are showing our new iPhone/iPad Application called ISEview.

  This should be in the Apple iTunes store in the next few weeks.  Well, as soon as we can get our developer to stop “tweaking/adding” cool things to it. 🙂   

Today this app is more useful in showing what can be done when you move to an open system, RESTful API.  In talking with the developers and marketing, we have some really cool things we want to add to this as we move forward.  As you can see from above, we can view ISE performance statistics as well as look at the MRC’s (Managed Reliability Controllers) and DataPac, power supply and cache battery health as well as some log information.  I can’t wait to see what other things Karl Morgan and Todd Burkey will be adding to it later!!

While the iPhone/iPad app is a “Geeky cool thing to do” again it speaks to what RESTful can do.  In our SNW booth we are also showing some work that Olocity did in their StorageIM monitoring product. 

My understanding is using our RESTful API SDK they were able to add our Emprise 5000 into their monitoring suite in only a few days.

Anyway, over the next few days I hope to catch up on some blogging, and also post a summary of “what I learned at SNW” so stay tuned.  BUT, no promises J

@StorageTexan <–Follow me on twitter!!

PS – thanks for @SFoskett for video taping a demo of the iPad / iPhone Application user interface. 

Who would have thought such a simple question – really more of “seeking to understand” as my VP of Sales Mark Glasgow calls it – would kick off such a slew of e-mails, comments and tweets. The other day I asked the following question on Twitter:

            @StorageTexan: question 4 all NFS/NTAP/Celerra Luving ppl. wht is ur opinion on y NFS is superior ovr block device access in #VSphere?

I got a TON of feedback, so much so that I decided to ask the question on my blog as a means to collect the responses and to allow others more than 140 characters to respond 🙂  Out of that came some really interesting responses, including George Crump sharing his views in Information Week.

 Let’s take a step back. The majority of my storage life has been in the block-based access protocol.  For the majority of my career it was all about Fibre Channel connectivity.  Have you heard the saying “If all you sell are hammers, then everything in the world is a nail”?  That’s sort of where I was at 5 or so years ago.  The FC was my hammer, and all connectivity issues could be resolved with FC!!  Then a few years ago iSCSI started to inch its way into the discussion.  Xiotech adopted it as another option for connectivity and I got to add another tool into my bag 🙂 

Then I asked the above question as a means of self reflection.  Why would someone choose to bypass block-based connectivity in lieu of file-based?  Just didn’t seem logical to me.  I even happened to have that particular tool in my bag, so it’s wasn’t as if I was trying to compete against it, just wanted to see what the big deal was.  Today, almost all storage vendors (Xiotech included) offer NFS connectivity.  Some utilize gateways like the EMC Celerra, NetApp V-Series as well as Xiotech.  Others use native support like PillarData and NetApp FAS product line. 

At first blush I thought, it has to be because IP is viewed as being less expensive, less complicated then native Fibre. But I think at this point, the price argument against FC should be put to bed, thanks in large part for Cisco/Brocade/Qlogic driving down the costs.  Complexity is also something I think should be/could be put to bed.  Just sit in front of a Cisco Ethernet and a Cisco MDS switch, the IOS is the same; the perceived complexity around FC is really no longer an issue.  Now, for the smallest of the SMB, maybe costs and perceived complexity is enough to choose NFS.  I can see that.

Maybe it’s because these gateway devices offer something their block based architecture can’t support.  That starts to make sense.  Maybe it’s some sort of feature that drives this decision.   In some cases, maybe its thin provisioning, better integrated snapshot, single instance storage/Data DeDupe and even advanced async replication.  Most storage arrays on the surface can do these, but with a gateway device maybe they can do this better, cheaper, faster, etc.?  For the SME/SMB I can see this as a reason.

Then again, according to some of the people that responded to my blog and twitter, maybe it’s for performance reasons.  Some sort of ability to cache the front-end writes make the applications/hypervisors/OSes just run quicker. Others suggested that gateway devices made it just “stupid simple” to add more VM’s because you could essentially treat an NFS mount as a file/folder, and you can just keep dropping VMDK’s (files essentially) into these folders for easier management.  That makes sense as well.  I can see that, I mean if you look at a NAS device it’s essentially a server that runs an OS with a file system that connects to DAS/JBOD/SBOD/StorageArray on the backend right?   It could be viewed as a caching engine.

Then it dawned on me, it’s not really about one being better than the other, it’s more about choices.  That’s what “Unified Storage” is all about, the ability to add more tools to your bag to help solve your needs.  If you look inside a datacenter today, most companies have internally tiered their applications/servers to some extent.  Not everything is run on the same hardware, software etc.  You pick the right solution, for the right application.  Unified Storage is the ability to choose the right storage connectivity for the various different applications/hypervisors and operating systems.  The line gets really blurred as gateway devices get more advanced and better integrated. 

Either way, everyone seems to be moving more and more to the Unified Storage device.  It should be interesting to see what sort of things come out of Storage Networking World in a few weeks !!

@StorageTexan

 The Debate – Why NFS vs Block Access for OS/Applications?

I’m going to see if I can use my blog to generate a discussion. First and foremost, I don’t want this to be a competitive “ours is better than yours”.  So let’s keep it civil!!

Specifically I’m trying to wrap my head around why someone would choose to go NFS instead of doing block based connectivity for things like VSPhere?  For instance, NetApp supports NFS and block based FC connectivity.  EMC typically uses CX behind their NS gateway devices.  Most Storage companies (Xiotech included) offer a NAS solution – either native or gateway. 

So, why would I choose to run for instance VMware Vsphere (as an example) on NFS when I could just as easily run it without ? Is it that the file system used for that particular companies NAS solution offers something that their block based solution can’t?  ie) thin provisioning, Native DeDupe, Replication etc.  Is it used more as some sort of caching mechanism and gives better performance?  Is it more fundamental and it’s more of a connectivity choice (IP vs FC vs NFS Mount)?

Please feel free to post your responses.

Thanks,

@StorageTexan

Truth, Lies  and Software Licensing.

Nothing get’s vendors more fired up then calling their baby ugly, especially if their baby is the mainstay of their product.  Case in point, there was a recent discussion around the importance, or lack thereof, in regards to automated tiered storage (A-T-S).  Jon Toigo jumped into the mix last week with some pretty good points.  Jon mentioned that he received positioning statements from a few vendors and twitter was going crazy with comments from all sorts of people weighing in on the subject.   Pete Selin over on his blog jumped into it with John Dias of Compellent.  So, to say that enough people have voiced their opinion on how, where and when “automated tiered storage” should happen, is an understatement.  Don’t get me wrong, StorageTexan is not immune to piling on, but I’m going to refrain.  Well, I’m going to try really hard to refrain 🙂

If we take a step back and look at this as just another feature in a storage array we then need to ask “what’s the value” of that feature?  In other words, why does a customer choose to go that route?  Most of the time features like this are positioned to reduce some sort of costs.  Thin Provisioning is positioned to allow you to under purchase storage, and then as you need it purchase the storage later at a reduced cost.   The theory is disk drive prices are always going down.  In the case of A-T-S it’s to reduce the costs of “Tier 1” storage.  Why spend money on storing your stale data on 15k RPM drives when you can move less accessed data down to 10k, or even 7.5k RPM spinning “rust” (shout out to Toigo).  At least that’s how I’m seeing this positioned in accounts.  Again on paper both of these 2 solutions make perfect sense.  If all things were equal, it is cheaper to purchase drives down the road (when prices drop) or purchase 7.5k over 15k disk drives.   The great equalizer is the cost to license these features.  Not just the costs, but software maintenance that goes along with them as well.  It would be easy to debate these features on their cost savings model if they would just charge you a onetime fee for this feature.  Unfortunately in most cases it’s a little more difficult.  Typically these types of features are licensed by capacity (Per TB charge), or in other case it’s licensed by disk drives or BOTH :).  So, when you add more capacity or more spindles, you licensed those drives or capacity for those features.  You are always hit with a moving target.   When weighing the pros and cons it’s really important that you don’t get focused (by the vendor) in only one direction.  In the Thin Provisioning example, the cost savings of not having to purchase more storage then you need might be outweighed by the fact that the license isn’t free, or that you have the potential of getting locked into that vendor for all future storage purchases.  In the case of A-T-S, not having to purchase 15k drives is offset by the software licenses needed to support that feature.  You might find that running everything in Tier 1 (or even Tier 2) vs. the licensing costs (Maintenance included)of automated tiered storage is a wash.

My advice, do your homework.  As I’ve discussed in prior blogs, don’t focus on the easy stuff.  Cost per RAW TB is not a good way to measure value.  Fully understanding vendor’s storage utilization best practice is important as well.  Understanding how their features are licensed and the software maintenance associated with those features are just as important.   I always like to ask my prospects, when comparing vendors, to not just get the upfront storage purchase price, but also ask for a quote to add 20% more capacity (spindles, bays, licenses and maintenance).  This isn’t the perfect way to keep vendors honest, but it does give you a good indication of what sort of cost savings you may, or may not see in future purchases.  Lest any vendor try to low ball on the initial purchase and then jack up the costs on the next one!!  Xiotech has a whitepaper called “Strategies for Measuring and Optimizing the Value of Your Storage Investments” that goes into some other thoughts on measuring the value of your storage purchase.  It’s worth a look.

Again, at the end of the day, do your homework and ask a lot of questions. 

 @StorageTexan