HP on the run

The Face of HP: Fear Uncertainty and Doubt

HP have just published a page on their web site called The Real Story about Cisco’s “One Giant Switch” view of the Datacenter.

What is written by HP on that page is so poor I expect them to pull it down or edit it soon, so for your enjoyment I’ve printed it to PDF for you to enjoy should they do so.

UPDATE 1:  A Cisco colleague pointed me at an “original” article by Paul Congdon, HP CTO, on Buyer beware of the “one giant switch” data center network model.  This has a bit more detail but not much.  In case that doc is pulled by HP, get the PDF.

I want to have some light-hearted fun addressing each of the points, but first I need to say two things:

1. I work for Cisco and these opinions, as declared elsewhere on my blog, are my own and not representative of my employer.
2. HP are clearly trying to scare “server administrators” that “network administrators” are going to steal their jobs, which makes HP into luddites who are afraid of change who feel more comfortable sharing that fear and not the kind of company I would want to partner with for future technology projects.

I should feel happy right now: this terrible HP article shows how little HP understands where their customers are, where their customers are going, and have little clue as to what their competitors are doing or what a detailed architecture looks like, especially a UCS implementation. Arguing over religious approaches to data center design, saying “my way is best, all others are rubbish” is not useful to anyone but your competitors.  I don’t see Cisco or IBM doing this, just HP.  I’m all for contrary and controversial points of view to spark debate, but the HP article isn’t in that category.

But is the reality (?) that this article is only representative of one part of HP and I have to believe that this is an isolated communication written by and individual that does not represent the rest of HPs data center and server business unit.  My confidence in HP is knocked, but perhaps this is the intention: a crazy article like this to throw the competition off the scent?  Or am I giving them too much credit now?  Oh, hang it, let’s get back to that crazy article!

Apart from trying to scare the server guys that the network guys are leering over the baffle boards and lusting after their keyboards, they make some biased, bogus, wrong… CRAZY statements… here they are, blow by blow:

Issue:  Ease of purchase/installation

Purchasing a Cisco UCS involves specifying a complex set of blade servers, blade chassis, UCS fabric extenders, UCS fabric interconnects, not to mention associated network and storage infrastructure…

Hmmm, this isn’t really complex at all – and not different from buying any other blade system, except all the components have been designed and built to work together unlike HP’s hodge-podge of home-grown blades and 3rd party switches.

I love this one.  Have you seen a bill of materials for a HP or any blade system?  Is it any more complex than buying eight 2RU rack mount servers?  What if the blade system has been around for a few years and carry all that legacy baggage, like troublesome management agents that your have to turn off in case they bring down ESX…

Buying a UCS system is easy: Cisco make it easy; it will be getting easier and easier as Cisco listen and learn from customers (you know how John Chambers is: it’s all about the customer, and that’s a winning culture, HP!) and the installation is a doddle as long as you don’t think you can lift a chassis on your own!  My colleague has delivered many UCS installations and, in one day, he can install the whole kit AND train the customer staff!  EASY!

As for Nexus 1000v, that is completely optional and is something that HP don’t have so they feel bound to knock it.  Nexus 1000v rocks, as any experienced VMware administrator in an enterprise will tell you, and will be a great success in large scale virtual deployments: but HP don’t get it, and I’m not going to explain why or persuade them to get into that business.

It’s a shame HP didn’t read this post and look at the “ease of installation” at the back of this UCS chassis!

Issue: Complexity

In Cisco’s one-giant-switch model, all traffic must travel over a physical wire to a physical switch for every operation…

The text is a bit vague here and only mentions one point about inter-VM traffic having to leave the physical host on UCS, whereas that’s not required for a “distributed switch”.  This has nothing to do with blade systems and everything to do with vSphere switches, and again you can choose from three types of virtual switches to suit your solution: what’s the point, HP?  Is choice a bad thing, HP?  Is innovation a bad thing, HP?

So this point about UCS complexity isn’t even about UCS, it’s about the Nexus 1000v which HP doesn’t have.  Is that bitter lemons I can smell?

They haven’t seen the beauty of UCS server profiles and how NO CONFIGURATION IS REQUIRED for the “northbound” Fabric Extenders or Interconnects: the system automates everything for you, and everything the system does is exposed via XML.

Designing, deploying and administering blade systems is not trivial and Cisco will not try to fool you that it is, but they are trying to help make it easier by innovative, purposeful design instead of useless FUD slinging at competitors that doesn’t help you, the customer.

HP luddites smashing up new technology

Issue: Control

The Cisco UCS model gives the network admin unprecedented control of the datacenter…

Here’s HP’s FUD for the server guy: someone is after your job!  What complete and utter nonsense, and quite insulting to the legion of smart admins who are multi-talented in virtualization, server administration, networking and storage.

With UCS, the LAN and SAN configuration is done once then you walk away and those engineers are not needed on a day-to-day basis.  The server administrator is still the king of deploying the physical and virtual assets, but they now have a unique one-pane-of-glass that all three disciplines can share to see exactly how the integrated systems are configured.  This is akin to breaking down barriers between disciplines and reducing configuration errors and improving the service that IT provides to the business.

You can use RBAC to silo the server, network and storage guy but you can also use RBAC to share the roles and foster teamwork.  You have choice, control, new options.

You can integrate UCS with a Manager of Managers because the system has been completely designed and built to be open, using DMTF standards and XML APIs.

Issue: Security

In a Cisco UCS data center approach, security happens at the core of the network – making security a major concern.

I had to read this one twice.  Network security is a concern, is it?   The other comments are priceless:

This is an integrated solution, so I guess if you crack part of it, you crack all of it

(Trying to stop laughing) I guess the HP article is a marketing page, so vague sweeping comments without facts or data seem to be the order of the day… and it gets better:

All a [knowledgeable] hacker has to do to get into this UCS system is to hack into the [Cisco] switch, which controls the data flow and the data itself,” Desai said. “For some [sophisticated] hackers, this is not that
hard to do.

I love how [knowledgeable] is emphasized.

There is no comment on the HP blade design.  Do they NOT use switches?  Of course they do, so isn’t the “attack footprint” similar?  Isn’t the HP system an integrated system (the Matrix!) – of course it is!  So what is the point, apart from making me laugh out loud!

Issue: Scalability

Cisco’s centralized approach is similar to a legacy mainframe – everything must run on one massive component, which becomes an expensive, hard-to-scale bottleneck…

OK, it’s at this point that I start to wonder if the person(s) writing this article have done any due diligence and know anything about UCS nor about data center networks.  Again, it comes down to choice: Cisco are not forcing you to use one mega-switch in your data center but they do give you the choice to deploy less network gear and use less ports: is that a bad thing? If you think so, you can architect a different solution still using Cisco technology or a competitor based on which is the superior solution that is fit for purpose and a good price – Cisco plays nice with partners and competitors, we don’t think it’s either/or but do think that we’ve got a decent solution set.

In Rich Gore’s Cisco-on-Cisco presentation where he talks about the huge use of UCS in Cisco’s enormous data center footprint, their challenge is enormous amounts of NICs, cabling, ports, switches – and his big plus point about UCS is the drastic reduction in this physical networking complexity whilst increasing his agility at reconfiguring logical network configurations for an ever growing server farm.

Yet again, I’m baffled by this vague comment – can someone explain it to me?  I thought choice, scaling, reducing complexity and increasing agility for the server admins was a good thing?

Did you hear the one about HP and Vendor Lock-In?

Issue: Vendor Lock-in

A report stated , “UCS will not accept blade servers from other vendors like HP and IBM. Nor will the Cisco-developed blade server within UCS work in any other vendors’ data center unification or consolidation platform.”

Oh boy, HP really saved the best until last!  This is like the killer joke at the end of a great comedic performance: it has the house rolling around in the aisles with tears in their eyes, burning this last joke into their brains so they repeat it at work the next day and get everyone at work laughing too.

Can you put an IBM or Dell blade in a HP system?  Does anyone do that?  Are HP really saying that they don’t believe in vendor lock-in?

Summary: Fun to read, but not to be taken seriously

Thanks, HP, for publishing that paper.  Not only was it great fun to read, it didn’t take long either because there was no detail in it.

If I was working in HP’s data center and server business, I’d be asking for that page to be pulled from hp.com and quick, and if I was a HP guy going to see customers over the next few weeks I would be praying that none of my customers mentions That Article.

HP’s “One Great Switch” is surely “One Great Own Goal”.

Related posts:

1. Cisco 6100XP aint Switches like VMware ESX aint Linux