Monday, April 30, 2007

Hu Yoshida

Is one of the finest guys in this industry. I wish I could be more like him.

I have known Hu since he hosted Ellen Lary and I on a whirlwind (and sometimes riotously humorous) tour of Japan and a series of meetings between DEC and Hitachi storage executives back in the early 1990's. He was fairly early on in his career then, but was so poised, articulate, and personable during that visit that it was easy to see he would one day become an industry leader, and he has...

Hu doesn't have a deceitful bone in his body. The terms Hu Yoshida and smarmy have never before been used in the same sentence until this one.

In fact, I am going to create and maintain a storage industry leader board with Hu is on the top rung. I bet he stays there for a long time.

BUT... (I have a sign somewhere at home that says, "everything before the BUT, is BS...")

Hu is wrong.

Virtualization does not just belong in the storage controller and we all know it.

Hu's not being deceitful, he's just not being clear. Virtualization as it relates to storage exists in at least 5 levels of the stack - the disks themselves, array controllers, intelligent switches, HBAs, and host based software. If I missed one or two, somebody will let me know. What I think Hu (as proxy for HDS) is trying to say that the virtualization of block mode storage should only be done in a purpose built storage controller (like Tag). Ok, that makes sense, depending on what we are virtualizing.

In fact, I would go further and perhaps state the obvious - maybe someday this obviousness will be called Kirby's Law - The initiation of virtualization is best accomplished at the layer directly above the virtualization target. (boy would I like to have a Law named after me...)

If you want to virtualize arrays, you put a virtualization layer directly above the arrays - not in HBAs or server software sitting two levels up in the plumbing. If you want to virtualize disks, you put virtualization on a storage controller not in an HBA, and so on...

Purists might argue that shades of grey exist - I myself have for years consistently argued that Tag is really an intelligent switch rather than a storage controller...and an HBA running RAID virtualization software, direct attached to a string of disks is actually a storage controller. Yeah, yeah...but stay focused on the bigger picture.

Kirby's Law (gawd, that sounds great, doesn't it?): The Initiation of Virtualization is Best Accomplished at the Layer Directly Above the Virtualization Target

If we take the above as true, the interesting discussion now becomes not where in the stack the virtualization resides , but what is being virtualized and how the virtualization is accomplished. Sustainable competitive advantage and real customer benefits can be derived from variances in endurance, sustainability, scale, performance, and other important qualities of the virtualization layer itself. More on that later...

Hu hereby enters the Storage Sanity Hall of Fame...and as you can all imagine...where there is a wall of fame, there must be a wall of...


Unknown said...

Kirby, first, I agree: Mr. Yoshida is best described as a "visionary gentleman." That is off the table.

However, I think you failed to identify for your readers why it's the "wrong" thing to do what HDS is proposing vis-a-vis storage controller level virtualization... On the contrary, I've talked to HDS customers who are virtualizing production data behind the USP and they are happy as clams. They have heterogeneous storage environments and complex migration/consolidation plans. It works for them. Is it "conventional" versus other block-level virtualization techniques in the network, in an appliance, or at the host layer? No. Does it work? Yes. We're not talking manned-mission-to-Mars here... we're talking data storage. There are many ways to thread the needle.

I think this is one of those topics that will infuriate people for years to come for the simple reason that many approaches will co-exist. I guess we all need rawhide bones to chew on (us analysts most of all, right?), but I think if you're going to call a visionary guy like Hu out on the carpet as "wrong", you gotta bring the ammo, dude. What don't you like? I am sure you'll never hear from them, but I'd guess that HDS would argue that they satisfy "Kirby's Law"... They use an advanced switch-based caching architecture in a controller that resides at the edge of the storage network to virtualize both internal and external LUNs... that is still "above" the virtualized target, per your law, under any circumstance. There are trade-offs in flexibility, but it doesn't violate your proposed law.

Good fodder


AceSage said...

he's wrong 1) using the generic term storage virtualization when he/they mean block virtualization 2) in stating that storage virtualization "belongs" in the controller.

My point is exactly yours, that many approaches co-exist.

My take on the TAG is that its two, two, two mints in one. Its a disk controller that does RAID (which is disk virtualization), and its an intelligent block mode switch capable of virtualizing RAID arrays behind it (same as SVC, StorageApps, Falconstor, Incipient, and a dozen others).

When all you have to sell is a block mode virtualizer, I guess everything looks like a block, and anyone who sees it different is a "fool"

My point is that virtualization of storage infrastructures makes sense on many levels, many aspects of storage are virtualizable, and "storage virtualization" belongs right above the element being virtualized.

AceSage said...

Today searchstorage reports that Yotta Yotta won a virtualiztion deal at AOL. left too soon!

According to the article, the customer bypassed pooling SAN storage behind its HDS USP. "The idea of using an 'edge' device -- whether a storage array or host -- to do virtualization to us seems a little out of whack," Pollack said. "You're adding a bunch of workload where a lot of data workload is already going to begin with. You need that array to become the pass-through environment for the storage it's already fronting."

Kirby's Law in action!!! said...

Hello Kirby thanks for the Compliments. I have fond memories of our Japan trip with you and Ellen Lary.

Please link to my blog to see my response