How Do People Get To High End Enterprise Storage?

One of EMC's core storage strengths is in high-end enterprise-class storage, as best exemplified by the Symmetrix VMAX.

But, in discussions with others in the industry, I've noticed a persistent "I just don't understand it" when it comes to this particular topic.  They openly wonder why *anyone* would need a device that powerful and that sophisticated.

For those of us at EMC, it's quite the opposite.  We meet these customers every day, and have been working closely with them for many years, sometimes for more than a decade.  For us, it's rather obvious why people continue to demand devices of this sort, and why interest is picking up.

So, in a non-salesy sort of way, allow me to attempt to bridge the gap?

A Few Definitional Terms?

Since we have very imprecise categories in our little storage world, allow me to describe the high-end enterprise storage category a bit.

From a technology perspective, we're talking multiple controllers (more than 2!) with large, dedicated non-volatile caches, multiple intelligent front ends, multiple intelligent back ends, all run with sophisticated and mature software and support processes.  These arrays are big, honkin' computers at their core.

There tends to be a heavy emphasis on block protocols and short I/O response time requirements, vs. file/object protocols and sheer bandwidth.  Predictable storage service level delivery -- no matter what -- is the core issue here above all else.

Got the basic picture?  Great ...

Scenario #1 -- The Application Is The Business

Very often, it's the case that one or more applications define the business: a trading app, a customer portal, a key database, and so on.  In these scenarios, there is no good excuse for a bad day.  

The serious financial and/or legal consequences associated with one or more applications being unavailable (or slow!) for any reason results in an understandable and logical business requirement for the "best available".  Whether we're talking mainframe, UNIX, Windows, iSeries or newer virtualized environments is rather irrelevant.

Conversely, if you don't live in this particular IT world, you'll find it difficult to understand and appreciate the rationale.  Any storage vendor who approaches this market had better be prepared to engage with some pretty seasoned IT professionals with several decades of large-scale IT experience.

Scenario #2 -- Little Arrays That Turned Into Big Headaches

Nothing wrong with smaller arrays.  EMC sells them, as do many other vendors.  They work great.

But there's a huge difference between having one or two of these devices, and an expanding fleet of 10, 20 or 100 that act as independent entities.  

First, there's a lot of tin that's not being used effectively in this scenario: cabinets, floor space, power supplies, processor boards -- and wasted capacity!  

Second, the team responsible for managing all of this has a growingly thankless task on their hands -- there's just too many thingies to look after.  Even though each array might not need much attention on an given day, there will be one or more arrays that have a problem, need an update, data migration, etc.

For an interesting presentation from Microsoft's IT group on this (very common) scenario, please see this illustrative preso from SNW.

Scenario #3 -- Delivering Storage-As-A-Service

This one is becoming a more common scenario as of late.  

An IT organization (whether traditional IT group or service provider) realizes that creating storage service catalogs that others can easily consume, manage, etc. is the preferred approach to storage.  

They define the storage services they'd like to offer, the portfolio of consumption models they'll need to support, and then go looking for a standard "storage engine" to generate a wide catalog of those services at optimized cost and operational efficiency.

This sort of approach usually leads you to either one of our unifed storage platforms (at modest scale) or something like a VMAX (at slightly more impressive scale).  The linear scale-out properties of the VMAX are appealing here as well -- especially for more demanding storage environments.  

Capacity, performance and functionality scales nicely and is implemented uniformly.  Associated costs and complexity either stays relatively flat - or more frequently declines - from a unit cost-to-serve perspective.

This Is Not About Vendor Olympics

Every product and every vendor has its strengths and weaknesses.  Competition is good for us, and our customers.  No arguments here!

However, those of us who work in and around these high-end environments have learned to develop a great appreciation about what makes them different from the more commonplace storage environments that are discussed so frequently.

Viva la difference!

Comments

[SANGeek]

Dear Chuck-
Another great blog from you. I love the slides from Microsoft. I think this really hits home in regards to consolidation. We recently went from several of vendor X arrays to 4 of vendor "E"s arrays in our consolidation projects. Overall we have about 1.5PB of which about 1.3PB reside on our NS-480 unified arrays and our other CX hardware and the rest on DMX. While we're only about 10% of what Microsoft has, I can see that as we scale out, storage is not going away. At some point the scalability/reliability of VMAX or Tier I hardware is going to be a huge advantage as growth rates emerge. My question to you is, we the CX4-960 approaching 2PB raw, and VMAX also having the similar capacity, plus the acquisition of Isilon, it seems like, as a customer we're gearing up for exciting times. Throw in FAST, Flash Cache, and things are starting to get easier. Again, throw in Vplex and Federation gets easier. At what point does Storage Virtualization, Cloud Storage and Tier I-IV converge into something I can just sit back, relax and let the storage manage itself? Is EMC working on some secret AI that it's customers aren't aware of yet? If so, keep up the good work! I think we will get there eventually.

[Chuck Hollis]

Hi SANGeek

Good to hear from you as always. And I'm glad that you're experiencing success with your EMC environment.

What you describe is what we call internally the "storage utility" - a set-and-forget model where you care absolutely nothing about the details, only the results.

Your timing is prescient in some ways. Before too long, you'll see us talking about some newer products that embrace this model, although not on your scale.

Our hope is that these concepts prove useful in the real world and at modest scale. If that's the case, we'll work hard to make them bigger and more encompassing of more complex requirements.

Thanks

- Chuck