Game Over In The CPU Wars?

I don't know how many of us paid any particular notice to Intel's recent announcement of the E7 processors.  I mean, isn't it more of the faster, bigger, more efficient, more functional same?

Well -- yes and no.

For me, it's the final sign that -- yes -- the era of Big Unix on proprietary RISC processors is finally coming to a close.

No Tears Lost

I came of age in the era of Big Unix, and its associated RISC microprocessors.  The 88000. MIPS.  SPARC.  PowerPC.  Itanium.  Long debates on who had the "best" processor -- fastest, most efficient, most scalable, etc. 

All apparently now irrelevant, consigned to the growing dustbin of historical computer science and technology debates. 

The answer is in: Intel wins, and so do vendors *and * customers.

The Complexity Tax

Having to support multiple versions of anything -- processors, operating systems, networks, etc. -- costs money. 

Speaking from a vendor's perspective (EMC) we spend an enormous amount of money each and every year to support the vast and diverse technology choices our customers make.  While that is laudable in many regards, the reality is that we could be spending all that money on creating cool new technologies, or deeper integrations with fewer choices.

The same thing plays out in our customer environments -- so much of every IT dollar gets sucked up in supporting diversity and its resulting complexity.  At some point, the technologist's cry of "best of breed" gets subsumed by the more pragmatic "we've got a business to run here".

Any time the IT industry can largely agree on a foundational technology (like Ethernet, for example) we all win: customers and vendors alike.

Early Cloud Discussions

Way back in early 2009 when I started writing about private clouds and all of that, one of the foundational tenets was that we were going to be living in an Intel world, processor-wise.  And that was going to be a good thing.

We had to be very clear at the time -- don't make the mistake of thinking that there will be meaningful SPARC-based clouds (still a viable processor at the time!), or PowerPC-based clouds, or (heaven forbid) Itanium-based clouds. 

Whether whether it was a public, private or hybrid cloud model being discussed -- the vast mainstream would be running on Intel, using presumably VMware as a hypervisor.  Two years later, that's exactly where we are.

The hardest part -- at the time -- was putting this rather controversial view of the near-term future in front of customers who were heavily committed to non-Intel processors.  They would push back, and demand a cloud approach that used their preferred combinations of processor technologies.

"Why can't you simply cloudify everything I already own?", they would ask.  Probably for the same reason we can't Ferrari-ize all the cars in your garage, I would think.

The enterprise computing and service provider world was going to Intel, and very quickly -- compared to other industry transitions.  The forces at play in the industry were just too strong.  Betting against Intel's preferred role would be like betting against Ethernet, I would claim.

The IDC Data Doesn't Lie

Those of you in the storage business probably recall that many of us consider IDC's market data the "gold standard" in the storage business.  Not surprisingly, they do the same thing for the server market as well.

Getting access to their data requires a paid subscription, but the message has been brutally clear for many years: every year, more money gets spent on Intel-based architectures, and less money gets spent on everything else put together.

Indsutry watchers realize that we're perhaps far beyond the "tipping point" where ecosystem effects take over.  Intel gets the amazing benefits of scale.  Everyone who writes software, or designs compatible hardware will inevitably Intel first and foremost.

Even here at EMC, that discussion has come and gone.  We've long been an Intel-first shop -- not only for our IT environment, but for most every hardware product EMC builds, and every software product we ship. 

Consolidation and standardization occurs, and then we all move on.  And that's what has clearly happened here.

That One Guy In The Room

I do a lot of customer briefings, and I occasionally meet That One Guy who isn't having any of it. 

I present my case for an Intel-centric world, based on virtualization and cloud principles.  And That One Guy digs in with both feet, and musters every argument he can to show that I'm either completely wrong, or at least FUD up the discussion a bit.

Frankly, I don't have to give this person the stink eye.  Why?  All his co-workers are doing it for me.  They're usually thinking "this is a losing battle, why can't we just move on?". 

My point exactly.

From Integration To Migration To Replatforming

If I was having a "we're going to a cloud model" discussion with customers last year at this time, there would likely be the inevitable discussion around how to incorporate and integrate their legacy application stacks (code plus processor) into their envisioned "cloud".

These discussions were awkward, to say the least.  I kept having mental images of trying to integrate fibre optic interfaces into punched card readers.

Fortunately, I haven't had one of those discussions in a long while.

Instead, the discussion is instead about migrating their critical applications to the new cloud stack: Intel, VMware, etc. -- most often in a Vblock configuration.  It's the new "big Unix".  And, for some customers, it's the new mainframe.

And, in turn, these early migration discussions have further evolved into replatforming discussions.  Migrations usually involve recreating the exact same functionality in the new environment.  Replatforming usually involves doing more with the new platform at hand.

Many of these larger, legacy applications try to solve the world's problems using their internal logic.  They try to manage their own performance, their own security, their own data protection, and so on.  Much of the interesting bits of re-platforming involves externalizing these requirements to be provided by the environment vs. from within the application.

Ultimately, their goal is to re-invest their "cloud dividend" in creating new functionality for the enterprise. 

And that's a very cool thing when you see it.

The Chain Of Good Deeds?

Intel continually doubles down in its architecture investments, and ends up creating the de-facto platform for enterprise computing.  The IT ecosystem recognizes this, and invests in creating value-add around the Intel platform vs. trying to support several flavors of CPU.

Customers recognize this, and gain enormous benefits in performance, efficiency, functionality and standardization.  The dividends are re-invested back into IT to create applications and infrastructure that can power the business in new ways that weren't acheivable in the previous model.

Has Your IT Team Gotten The Message?

If you're running a decent-sized IT shop, look around and ask the question: what's the plan for migrating off the niche stuff, and getting onto the mainstream?  The reason is clear: with every passing year, you'll find yourself at a growing disadvantage compared to those that have already made the move.

You'll be paying more, and getting less.

Internally, you'll probably have a tough road.  For example, you'll likely encounter Stockholm Syndrome in the process.  Many of your IT staff have likely built their careers -- indeed, their self-identity -- around these proprietary processor and software stacks.  They spend lots of time with the vendor.  They go to the specialty shows and events.  They network with others like them. 

And no amount of logic is likely to budge IT people in this unfortunate situation.

Now, to be fair, the vendors of these non-Intel server systems will likely move heaven and earth to get you to stay with them.  They'll throw in discounts, professional services, free software -- whatever it takes to get you to not make the big switch to mainstream server and software architectures.

Don't underestimate the persuasive power of a large, entrenched vendor.  IBM, in particular, is very proficient at this particular skill.  For example, to this day you'll still meet people running Lotus Notes (or i Series!) for example.

Maybe you don't have the resources right now for a full-scale migration, but there's no real excuse for not having a stated point of view, and a rough plan to get there.  And there are plenty of capable professional services firms (EMC included) who'd be glad to help you get from here to there when you're ready.

Back To The Role Of IT?

I believe one of the most powerful things going on in the IT world is the shift in roles: from being viewed as a cost center to being viewed as a value generator.  Ruthlessly standardize and optimize how IT is built, operated and consumed (e.g. cloud) -- and invest your new-found "cloud dividend" in becoming more relevant to the business -- more agile, more responsive and more innovative.

And, from that point of view, being able to standardize on a single enterprise processor architecture represents yet another opportunity to pursue that worthwhile goal.

Comments

[EwanToo]

A good post, and I agree with pretty much all of it, but it's interesting to see just as we reach the end of "big iron" processors, at least in volume sales, we're seeing the new introduction of micro-servers running low power (and pretty slow) ARM chips.

Whether ARM makes a dent in the Intel armour, I'm not so sure, but it's an interesting comparison to Intel x86 vs Sparc and Power in the 80s and 90s.

[Massimo Re Ferre']

All true.

Amen to that.

Massimo.

[Chuck Hollis]

EwanToo

Agreed. There's no "end game" in all of this. And we're all watching for the next wave to form, if it does.

-- Chuck

[Iso Belgesi]

Interesting point, i've never think about it like that..