There's a news story today about Dell, IBM and HP all working on or having released new blade servers. Huh? What's the point of those at this point?
You can easily (easily) use up all the power and air conditioning available in most facilities with "regular" servers. Why pay more and get a proprietary form factor etc? Are there enough corporate customers who have plenty AC and power available in a small computer room? Or is it something else?
Data center space is an issue in many places. We have several thousand boxes and we pay a huge amount of money just for the space.
Do blades use less power per some performance metric than "regular" servers? I have no idea if they do, but could see a "green" argument to be made if they could do the same job with less power.
If the blades are *more* expensive, then you are doing something wrong. One of the compelling reasons to go with blades is that they are cheaper than "regular" servers.
As to corporate customers with plenty of power in a small computer room, I previously worked at one company with lots of servers (30,000+) where consolidation to smaller hardware with lower power requirements actually allowed them to consolidate enough to sell off an entire multi-million dollar data center, saving millions of dollars.
Everything is relative. The blade servers are more compact than full-size servers, so if your most constrained resource is floorspace, the blade form factor gives your data center more capacity.
But, naturally, once you eliminate one bottleneck, the next most constrained resource forms the new bottleneck. So power and AC may become the limiting resource as devices become more compact.
I think there we will see increasing demand toward making virtualized servers with enough stability and performance to substitute for physical production servers. You do need a very beefy server to run a lot of VM's, but the ratio between processing power to electricity requirements tends to improve.
VMware claims that their product pays for itself in only about 12 months of operation, given the savings in power and AC after converting a bunch of physical servers to VM's. YMMV, of course, but it's an interesting way of looking at it.
The answer is yeah, there's plenty of places where there's plenty of AC and power available to permit them to go for higher server density than standard 1U servers can provide them.
Especially if you look at places who have ample chilled-water available on site (as many places which used to service mainframes do), you can get into really convenient cooling solutions that way.
The blade chassis can use power and cooling more efficiently.
From what I remember when HP was pitching the c-class chassis to us, instead of running 4 power supplies at 25% each, you would run one at 100% and the rest as standby. Well, probably not 100%, but at least run as many of the power supplies at peak efficiency as possible. The fewer AC power supplies you run, the less energy you lose to the ac/dc conversion.
You also save a ton of cable management effort -- they typically have a switching backplane built in so you don't have to have 20+ network cables, fiberchannel and power cables for 20 servers. It's more like 4 power cables, 4 fiber channel and 4 network cables. That also could mean fewer switches, which also means less power and cooling.
In those situations where density is not the objective, there are still many compelling advantages to blades: shared components (six power supplies for dozens of hosts), reduced cabling (network equipment built into the rack), faster deployment (many companies will deliver a rack of blades already hooked up and configured - just add power and net), host upgrades without unplugging and plugging in cables and other economies of scale.
If you knew you needed to add 1000+ hosts to your data center as fast as possible, you would seriously consider blades. It's not for everyone, but for some it's the only thing.
colocation companies charge per U of space, a blade server solution full of blades would give you a better bang for the buck.