I agree. I don't know why Liu made that assumption. Most large scale data center operators agree that blades are not a good solution:
1. Not cost effective because they don't benefit from the economy of scale of commodity hardware like pizza box servers.
2. More expensive to operate as they often require redesigning/reinforcing cooling as they concentrate too much power in too little space.
3. The space-saving advantage of blades doesn't address the root of the problem: any data center architect worth his salt will design a data center to be power-constrained, not space-constrained.
1. Not cost effective because they don't benefit from the economy of scale of commodity hardware like pizza box servers.
2. More expensive to operate as they often require redesigning/reinforcing cooling as they concentrate too much power in too little space.
3. The space-saving advantage of blades doesn't address the root of the problem: any data center architect worth his salt will design a data center to be power-constrained, not space-constrained.
Liu should have read his source more extensively, James Hamilton, VP at Amazon Web Services, who has criticized blades for these exact reasons: http://perspectives.mvdirona.com/2008/09/11/WhyBladeServersA...