Game on! The rapid rise of bundled infrastructure products

Although I work for 3PAR, which is an EMC competitor, I admit there is something very intriguing about the vBlock concept - and that is the concept of buying pre-fabricated, pre-integrated, information systems. Its like the old LAN in a can or SAN in a can, but vBlock is a much more sophisticated bundle with a much broader vision than previous network bundles.  Integration is the dynamo of technology after all and maybe years from now, maybe we'll see Intel come out with a data center on a chip! (rim shot and cymbal)

But seriously folks,  the technology barriers to entry in this "data center in a can" business are not all that high. HP is planning on announcing their reaction today and  I expect that we'll see similar concepts announced within 6 months from various vendors and other industry alliances.  The stakes are too high for server companies to ignore. FWIW, if you haven't read Chris Mellor's excellent article in the Register from yesterday about HP's Data Center Cells, you should.  

To underscore how low the barriers to entry are, think about how easily Cisco, a company with no server sales to date, is using vBlock to legitimize their UCS server business.  By burying UCS in the data center bundle, Cisco can claim certain things that are unique and special about it's server business, but I doubt the rest of the server industry will grant them (or HP) uniqueness for very long.  The only thing keeping IBM from doing something similar is IBM itself.  Certainly they could draft people from IGS to respond very quickly if they wanted to.

As we see vBlock knock-offs emerge, its going to become increasingly obvious that the integration of all this virtualization technology is not that difficult when you use cloud-oriented virtualization technology.  The sudden change of positioning from EMC, Cisco and VMware is curious - for many months, virtualization has been the thing that makes life easier and all of a sudden now we are to believe it it is the thing that needs to be made easier. 

Infrastructures by nature are adopted and used based on their stability (which includes scalability), flexibility, and efficiency.  The integration characteristics and features of individual products in those infrastructures will determine how successful those competing infrastructures are.  For both servers and storage integration, these characteristics and features will include:

• robust, modular components that can scale from small implementations to large ones
• efficient provisioning and re-provisioning of resources - including storage reclamation
  
• high densities, such as VM/server ratios and storage utilization
• predictable performance for mixed workloads and under stress conditions  

FWIW, professional services are a drag on efficiency as added cost or a drag on flexibility as added time to deployment.  Ease of use and automation features will continue to be an important key for cloud vendor success.

I used a fox in the henhouse photo the other day, when I posted in reaction to the rumor of Cisco and EMC competing with cloud service providers. And yes, I probably deserved much worse than the reaction I received from both Chad Sakak (The Virtual Geek) and Chuck Hollis. 

Of course , the most clever fox eats eggs instead of chickens.  The problem with professional services engagements is that they have a way of sticking around a lot longer time than expected.  It's entirely possible that vBlock won't require any, but it certainly looks like Cisco and EMC want to sell them.  I have to hand it to both EMC and Cisco - they are very clever companies.

Comments

[nate]

I think the price tag of vBlocks ranging from $1M to $6M starting price will make those who buy it struggle to compete against the Microsoft's, the Google's, the Amazon's of the world and other cloud players. When your coughing up $6M for 64 blades and a V-MAX that's just alarmingly obscene. Sure there will be discounts but still..

Web pricing for 4 HP cClass blade chassis with 192 cores each and 1TB of memory runs about $600k(that's 64 blades). That is online web pricing, no talking direct to HP no special discounts or whatever, that's what I can buy it with right now on my credit card..if I had a card that had a limit that high anyways.

That's with 10GbE virtualconnect and 4Gb fiber virtual connect(I don't need 8Gb). I'll take that over FCoE any day.

Toss in some 3PAR storage on top of that, I mean it's almost comical what they are advertising the pricing at.

posted more on this on my blog last night http://www.techopsguys.com/2009/11/03/the-new-ciscoemcvmware-alliance-the-vblock/

The whole integration thing to me is a joke at this level. I mean your talking about 64 servers, paying that high of a price I don't care if the thing is able to drive itself to my data center and wire itself automatically. ITS ONLY 64 SERVERS. They must be selling stuff to inbred monkeys to need to pay that much for hand holding. With no integration at all we can have a new vmware box built and configured in as little as 30 minutes(about half of that is configuring vmware itself).

Now if your talking about hundreds or thousands maybe it makes more sense, but also consider the advertised price tag of those 64-server vBlocks starting at $6M, then ask yourself how important is that tight integration really worth?

And I bet if it was a large number of servers I could nail something down with the HP cClass and virtualconnect with the MAC/WWN dynamic mapping and 3PAR virtual copy to configure the systems just as fast, if not faster(after spending a bit of time to get the first iteration of it operational). Right now I spend 30 minutes because I can spend 30 minutes, it's not a big deal.

I'd rather have more servers, more storage, and take the rest in a higher paycheck thank you very much.

It's a scam, and I feel sorry for anybody who falls for it, oh wait, no I don't. It just makes people like me that much more valuable in the world.

[nate]

Wanted to clarify a bit that $600k price tag for the HP is for 64 blades, so a total of 768 real CPU cores(no hyperthreading), and 4TB of memory, which can fit in 1 rack(4 chassis = 40U total)

How many VMs you think that can hold? How much fancy management stuff are you willing to pay for to support 1 rack of equipment? Keep in mind much of the management beyond the hardware level is all handled in vmware land.

At one point I heard a claim that VMware's average customer gets something like 8 VMs per core, so that equates to 6144 VMs.

[marc farley]

Great blog post, analyzing the costs of vBlock. We are always looking for ways to make provisioning easier and we'll have some new news on that shortly - but you probably know about that already. :)