Converged solutions are simply not simple

Complicated infrastructures make data center operations difficult, but they don't start out that way. The solution to fixing the infrastructure might work but might be equally complicated.

This article can also be found in the Premium Editorial Download: Modern Infrastructure: Production workloads go boldly to the cloud:

I've been thinking a lot about two areas of IT: complexity and convergence, which have a lot to do with each other.

Collectively, we operate some very complicated infrastructures, which make many aspects of IT difficult. With a massive funding round for Nutanix, the integration of Whiptail into Cisco as UCS Invicta, and VMware's release of VSAN, we are seeing a lot of convergence happening. Normally siloed areas of IT are being forcibly integrated.

Our data centers are piles of kludges

We don't start out with complicated systems. The first system designs we do to solve a problem are concise, simple and easy to implement.

Problems tend to arise during the implementation process. Sometimes it's a missed requirement--somebody didn't talk to everyone or the right people. Sometimes it's an assumption, perhaps that the management interface on your Fibre Channel switch could do gigabit speeds, but it's actually only 10/100. At any rate, you need a fix, and that fix makes a mess of your clean design.

Gradually we add things to a nice design to handle new business requirements. We replace aging or failed components with new parts, but those new parts are never the same as the old ones. We come up with numerous "easy fixes" for problems and they stack up, kludge upon kludge upon kludge, until all that's left of the original, simple system design is the Visio diagram from years before.

Converged is not simple

Convergence is seen as a way out of this. Buy a setup from Nutanix, and it has compute and storage integrated and managed as one. Same with VMware's VSAN and a pile of local disk drives in your servers. Buy a Cisco UCS, now with storage, and you get a single management interface for network, storage and compute. Very attractive, and on the surface, it's simple to own and operate.

These things are not actually simple, though. There is a lot of complexity hidden under the hood of a Nutanix system, a lot of moving parts and pieces, reminiscent of the mind-bending movie Inception: virtual services implemented by virtual machines running inside the environment built on these services.

The same goes for VMware VSAN, or any other highly converged solution. Cisco UCS takes a more traditional hardware approach and thus avoids some complexity. The vendors make up for it in other ways, though, particularly with their converged Fibre Channel and data networking. Any time you mix traffic, you create something that is more complicated and harder to manage, despite what any vendor tells you.

We like these converged systems because they simplify day-to-day operations and cut operational costs. But they really don't simplify our data centers. They actually add complexity, but try to hide it from us.

In 2002, U.S. Secretary of Defense Donald Rumsfeld spoke in the context of war about "known knowns, known unknowns and unknown unknowns." While the quote itself is complicated, his message is simple: Sometimes we don't know what we don't know, and that's bad. When we buy converged systems, are we inviting a lot of unknown unknowns into our data centers?

There isn't a good answer to this, and I actually like a lot of these converged systems for what they can do to bridge the gap between older, complicated, kludge-filled data centers and cleaner, simpler ones.

To me, they're black boxes that someday I will have to open, because someday something will go wrong. On that day, I will be longing for the nice, simple infrastructure we once had, one protocol to a cable, one workload to a server and a single diagram that shows it all.

This was last published in April 2014

Dig Deeper on Converged infrastructure (CI)

PRO+

Content

Find more PRO+ content and other member only offers, here.

Join the conversation

2 comments

Send me notifications when other members comment.

By submitting you agree to receive email from TechTarget and its partners. If you reside outside of the United States, you consent to having your personal data transferred to and processed in the United States. Privacy

Please create a username to comment.

Yes, there is a lot of complexity hidden under the Nutanix hood. And it stays hidden. That's the point.

There's no free lunch. There is complexity in any environment. The question is, "who takes responsibility for the complexity?" That's what a converged IT vendor should do. If done correctly, then life for the admin gets better.
Cancel
Bob, I disagree with lumping in Nutanix with UCS in terms of converged infrastructure complexity.

I've long been a huge UCS fan - and if you do a Google search on UCS vs. HP Matrix, my posts still come up at the top of the search rankings. UCS, as the first purpose-built platform for hosting virtual infrastructure, was very innovative in its day. But UCS has a tremendous disadvantage in that it only addresses the compute side of the virtualized datacenter - about 30%. By far the majority of challenges have to do with storage. This is why EMC, NetApp, Hitachi and Nimble all utilize UCS as the compute side of their converged infrastructure offerings.

And UCS does have some very sophisticated, albeit complex, capabilities. For example, the GUI enables the storage and server teams to more effectively collaborate. If the server team over provisions storage and is about to take a LUN off-line, the storage team can hopefully catch them and run down the hall to stop them.

Nutanix doesn’t have any LUNs to worry about. In fact, there are no separate storage guys – everything is managed by the virtualization team. We estimate that it takes 1-3 weeks to set up UCS and another 1-5 days per week to administer it. And, of course, this doesn’t include storage which tends to require equivalent set-up and administration times. In contrast, Nutanix, which includes both compute and storage, takes about 30 minutes to set-up and 1-3 hours per week to manage.

Internally at Nutanix we talk ceaselessly about being “uncompromisingly simple”. We strive for extraordinary simplicity in our products, our ordering, our channel programs – in all that we do. This simplicity is validated again and again not only in our customer accolades, but in our status as the fastest-growing infrastructure company of the last three decades.
Cancel

-ADS BY GOOGLE

SearchWindowsServer

SearchEnterpriseLinux

SearchServerVirtualization

SearchCloudComputing

Close