USP V Clustering - initial thoughts

Some very early thoughts on yesterdays clustered storage arrays announcement - Hitachi High Availability Manager, or HAM for short.

<this was written while watching the Champions League final and twittering last night so more info might have emerged since I wrote it>

Where I come from in the world, "ham" is pig meat.  Let's hope this doesn't turn out to be a technology that resembles slapping lipstick on a pig, in an attempt to make the pig more attractive ;-)

What it is and what its not

First up, the announcement is basically a new supported configuration  - clustering of two USP V's.  Seemingly they will use Synchronous TrueCopy  to achieve mirroring normally within a single campus to create a 2-way Active/Active cluster.  More than that I don’t know at the moment and even that is based on some assumptions.

Unlike the recent Symmetrix V-Max announcement, it is not an announcement of new hardware, internal architecture, or anything to do with feeds and speeds.  If you were expecting new hardware or major improvements to existing software features such as Dynamic Provisioning, Distance Replication, Snapshots etc then you might feel a little short changed, and clearly some people do.  In fact, my initial feeling was one of being somewhat underwhelmed, and that's coming from someone who knew that this was essentially to be a storage clustering announcement.  But even I was expecting a little more than what was suggested on the solitary diagram, used during the Webex.

Second up, there is as yet very little information in the wild on how this will actually work.  In fact, the below diagram was all that was provided on the call.  With this in mind, it's not possible to do any kind of deep dive....   And to be fair to those feeling short changed, looking at the diagram there doesn't look to be *too much* anything new from a hardware perspective.  Added functionality to the microcode is obviously required, but as previously mentioned, this is essentially a minor change.

Steady as she goes

However, and I think this is an important point.  HDS are claiming a major new feature(s) resulting from a minor change.  And from where I stand most days, this is a win win situation!

If I take a moment for a sidestep.  I recently mentioned that in my opinion the fact that EMC stuck with crown jewel of the Enginuity code base in the Symmetrix V-Max was an excellent move (too much change is too much risk).  At the top end of the market, where much of the USP V install base lives, making major changes to anything, including hardware or software, presents a risk.  And we all know how risk averse people at the top end of the market tend to be, and for good reason.  Therefore, achieving a major improvement via a minor change is "win win" - less development costs, no fork lift upgrade, and less perceived risk to the customer.  This is all of course assuming that some of the functionality achieved from clustering a pair of USP Vs will yield "big wins".  Only time will tell.

A final point for those feeling short changed (I did initially, and with the lack of technical detail available I still havent been able to decide if I got "value for the hype").  Take another second to think about the recent Symmetrix V-Max announcement (apologies to both parties for the parallels) and in particular - FAST.  Some the goodness that EMC FAST promises to bring, are worthy of a separate mid-life V-Max announcement.  In fact FAST was not announced as being available at V-Max 1.0. 

Think in particular about the promised ability of FAST to migrate/re-tier volumes that are actively being replicated (SRDF...).  Might seem small compared to all the new glitzy hardware on show in the V-Max.  However, from a business and customer point of view, this is a big win.  Consider that re-tiering products that require suspension of remote copy operations while the re-tiering operation is in progress, such as Hitachi Tiered Storage Manager, sit on the shelf in large TrueCopy and HUR environments and only have the dust blown off them for planned out of hours migrations.  They're not much use in production environments.  For FAST, online migrations during active remote replication load will be a big win to those kinds of customers.

Likewise, for some of the promises associated with USP V clustering (I can't get my head around calling it HAM).  Large scale Zero Downtime Migrations (ZDM) is just one ability that will be a huge win for customers and businesses that don’t like downtime - anybody know any customers and companies like that?  Actually, if I understand the possibilities correctly, the term "migration" may no longer be appropriate in clustered USP V shops.  I don’t know a customer that would not like that!

Will it cost the earth?

Of course, any potential will come at what looks like a huge cost. 

2 x USPV != cheap 

But as a general rule in life, you get what you pay for, and this is one that looks like it will be reassuringly expensive ;-)

On the topic of cost, I recently did some work that resulted in the purchase of new kit to the tune of slightly north of £2M (not huge but also not insignificant).  However, over 3 years this actually resulted in a substantial saving for the customer as they were able to scrap their "extortionate maintenance fees* tended maintenance fees.  USP V clustering and Zero Down Time Migrations makes it possible to never have to fall into extended maintenance on current and future HDS storage.

That said though, from where I'm standing, implementing such a USP V cluster looks like it might cost the earth - basically double almost everything in the USP V - cost, power and cooling.  So it might not just be a case of having extremely deep pockets, there is also the environmental impact.  Talk about difficult to justify in current climate! (pun intended)

In fact I recently sidelined the idea of some SSD in an array because a single 8 disk array group was going to cost £2,000 less than the first house I bought in 2002!  That was despite the fact that they would be far more environmentally friendly.  So doubling the cost of a new USP V might be hard to get past investment review boards :-S


Its difficult to really say much, due to the lack of technical information out there.  However, after some thought while watching Man Utd being schooled by Barcelona last night, I'm starting to see some potential, and feel less and less like this is a mid life crisis for the USP V. 

Of course, like most people, my view is from a distance.  Up close and personal might reveal it to be a pig with lipstick, but at the moment nobody knows enough to say that.  In fact on the contrary, I think it is fair to say that Hitachi has a heritage of bringing some great array based products to the table.  With this in mind, I think writing this off without more knowledge or experience is probably a little rash.

Oh and Zero Downtime Migrations is not the only thing HAM will bring to the table.

Final thought - who came up with the acronym "HAM".  Not great!  Whats next, naming TrueCopy/HUR for z/OS "TURKEY" -
Universal Replicator
for Count KEY Data volumes

PS.  Anyone with any technical info on the matter please feel free to contact me.

Nigel  (I only ever talk about storage, never what Im eating for breakfast)