Five Things Not Wrong With Blades
SearchDataCenter.com has a short paper (registration required) on the top 5 objections IT folks have to blade servers. I agree with the paper's discounting of all five, although a couple of them have partial merit.
Objection #1: Blades are a single point of failure.
The paper says there's no single point of failure if blades are configured correctly. For example, connecting blade enclosures to two different power feeds removes power supplies as a lynchpin. No, there are still single points of failure in blades; they just incur acceptably low risk. For example, the sheet metal side wall of a blade enclosure has no redundant backup, but sheet metal doesn't often spontaneously crack or disintegrate. (And if an enclosure wall cracks, it's unlikely to cause server downtime.)
Objection #2. Blades concentrate network connections beyond manageable limits.
Blade neophytes picture the rear of a blade enclosure as a rat's nest of cables. Nope, it's actually very clean. Connections are trunked, with network routing configured via a management app that scales well (like through Virtual Connect Manager or vSphere).
Objection #3: Blade servers generate more heat than rack servers.
Sometimes (but rarely) this is a valid concern. Per unit of compute, blades generate less heat than rack servers. However, since blades are concentrated into a smaller space, they can generate more heat per square foot. A 42U rack of 64 c-Class blades will likely produce more heat than the same rack filled with 40 traditional 1U ‘pizza box’ servers.
Objection #4: Blade servers aren't green.
As the paper says, that's false. OK, they’re not built entirely from rainbows and butterflies, but blades do consume less power, which means a cleaner footprint. One great example: The Scandinavian countries have the highest adoption rate of blade servers of any region in the world. Why? The customers and partners I spoke to at VMWorld-Copenhagen last week universally gave the same answer: Blades are green, and those countries are highly attuned to that.
Objection #5: Blade servers cost more than rack servers.
It depends on how they're configured, and what's included in the cost. At some point of around a dozen servers, cost for blades are often lower than 1U or 2U rack-mounts, but it's heavily dependent on configuration and networking costs. Here’s a tool that helps you compare costs of rack form factor versus blade form factor.
