slide

Dear future hci partner

Ned Bellavance
5 min read

Cover

If there’s one thing I wish HyperConverged Infrastructure (HCI) vendors would stop doing, it’s promising that the product will be up and running “in a matter of minutes”.  First of all, it’s simply untrue.  Second, it’s irresponsible and sets those of us deploying the hardware up for failure.  When skewed perceptions intersect meaty reality, the deployment engineer is the first to be skewered.  And you know what else?

I’m not speaking in a theoretical sense, but rather as someone who has deployed HCI solutions from four different vendors in the last six months.  I’m not going to name names here, but the issues are so consistent amongst the four horsemen of my personal HCI apocalypse that I feel confident extending this to all solutions currently on the market.

First of all, the technical presales and their marketecture has got to go.  Yes it is technically feasible to deploy your solution in 15 minutes, or rather it is possible to run the wizard in about 15 minutes.  But that requires that:

  • The network is configured correctly based on whatever strange requirements your solution has (native VLAN, IPv6, IGMP, BGP with ISIS over PVXLAN)
  • The technician has read your 600 page document, and the 75 page upgrade guide
  • The pre-deployment worksheet has been filled out, reviewed, updated, rejected, and resubmitted
  • You haven’t accidentally used a naming prefix that breaks the entire deployment and requires a full re-image (True Story!)
  • The equipment has been unboxed, racked, cabled, powered, and labeled
  • The 25GB of software updates have been downloaded
  • It’s not a full moon or a day ending in ‘y’

Then SURE, YES, YOUR WIZARD is AMAZING!

But it’s not.  The wizard is part of a UI that is explicitly designed to hide the complexity of the solution.  The problem is not the complexity.  I’m fine with a solution being complex, in fact I expect it.  Complexity is okay, hiding complexity badly is not.  The wizard and management GUI in most of these solutions is the front end to scripts that often break, fail, or give confusing feedback.  HCI is still in infancy, and I know that making complexity simple is actually very difficult.  If you want to see it done well, go log onto Azure or AWS.  They take an incredibly complex set of technologies and wrap a streamlined GUI around it.  It is in their best interest to make the consumption of their services as easy as possible, so they have every incentive to lower the barrier of entry on their platform.  If you want to use the full power of the platform, you can of course use the CLI or the APIs, but you don’t have to.  Azure, AWS, and other cloud platforms have had over a decade to streamline their interface and make sure it is intuitive and resilient.  HCI, well not so much.

If HCI solutions were truly so simple deploy that anyone could do it, they wouldn’t require a 600 page installation manual.  If the solution can truly be front-ended by a simple wizard and management GUI, then a manual that fells a small forest every time it is printed would not be necessary.  If you try to hide complexity, and do it badly, you are simply inviting poor deployments and disillusioned customers.

And don’t tell me about your 200 node deployment that went flawlessly at client X.  That deployment was designed and executed by a senior engineer from your company who probably wrote half the interface and scripts.  Of course someone with intimate knowledge of the platform had no trouble deploying their own software.  You know what a great test of software is?  Hand it over to someone else and have them run it without reading the source code.  They don’t know what assumptions you’ve made, or what you expect the user to do.  They can and will find bugs.  You know who is doing that testing? ME.

Also, let’s not talk about POCs.  A Proof of Concept is usually executed in a standalone environment that doesn’t mirror actual Production.  Most POCs are successful, because they only confirm that the solution works as already tested by the engineers at the vendor.  Look, I know your technology works, mostly, and that’s not what I am annoyed with.  The tech is not the problem, the marketing IS.  Remember, every time you run a successful POC, a baby Arctic seal is clubbed by Katy Perry.  And that makes her cry.  You don’t want to make Katy Perry cry do you?  You monster.

A few weeks ago I sat through yet another HCI presentation by a vendor wherein they started making claims about how easy it was to deploy their solution, and how it offered performance and features unparalleled by the competition.  Naturally, my colleague and I expressed some cynicism about the marketing hype.  The solution engineer tried to tell us that it wasn’t marketing hype, and things started to get a little heated.  I told him that we hear a lot of great promises, and when the marketing meets the road things begin to fall apart.  I’m not going to drink the kool-aid just cause you have some PowerPoint slides and a true believer on your side.  The thing is, I know this is marketecture BS.  The clients you are hocking your wares to don’t.

I’m sure your solution is a unique, precious snowflake, just like every other HCI solution put in front of me.  And I’m sure it works, well probably, I wouldn’t know since I am still reading through the 350 page administration manual for a solution you describe as dead-simple to deploy in a matter of minutes.

So please just stop it.  STOP.