Taken from the HP blog site:

Just saw an interesting article at http://searchservervirtualization.techtarget.com/originalContent/0,289142,sid94_gci1239548,00.html?track=NL-652&ad=577295&asrc=EM_NLN_919007&uid=5752851, written by Alex Barrett, News Director at Search Server Virtualization.com. I normally don’t like to quibble about product details or limitations, but this one had some serious misconceptions about blades as an overall architectural concept.

The substance of the article was reporting that a consultant, Rod Lucero, CTO of the VMware consultancy VMPowered, prefers to recommend standard rack-mount servers as opposed to blade servers for virtualization projects, predicated on several limitations he cited in the, including:

• Lack of I/O bandwidth, predicated on the assumption that multiple VMs supported on newer high-core count CPUs would overwhelm the I/O capabilities of the blade chassis.

• An assumption that the blades would be diskless (“Lucero's other gripe with blades is that they are often sold diskless, thus requiring the blades to boot from SAN – a less-than-perfect process”)

Now at one level, I suppose as an employee of HP I should not complain about his position. We are, after all, the leader in rack-mount servers, and if Mr. Lucero recommends rack-mount servers, I assume we’ll get our share of them.

But the Mr. Barrett and/or Mr. Lucero may have missed some of the more recent changes in the blade arena, and have certainly missed the element of choice and diversity.

On the I/O side, with all of the leading vendors now ramping up to 10 Gb Ethernet uplinks, along with fringe alternatives such as InfiniBand, only a very few application environments should represent a real challenge to the aggregate bandwidth of the chassis. The HP c-Class, for example, can be configured with as much as 28 Gb of Ethernet uplink capability. Even our arch-enemy, the Evil Empire in Armonk, has announced 10 Big uplink capabilities. Assuming 8 four-socket blades, that is 3+ Gb per blade, certainly adequate uplink capability for even a large collection of VMs.

On the disk/SAN side, Mr. Lucero is right, SAN boot is often a less than perfect process, and is not for everyone. On the other hand, a lot of IT shops are using it successfully. The key is choice. Nobody, not even the bad guys in the blue suits, white shirts and wingtips, will force you to buy diskless blades if you want them with disks. We love to sell blades with disks. We’re not philanthropists, we make money off them. The good news is that you can have them your way. Or both at the same time.

While there may be some legitimate environments where only rack servers will be the answer, I think that the overlap is now much larger than Mr. Lucero believes, and it would behoove him to take a closer look at the most recent developments.

Read or comment on the original, here.