ENKI - Managed Cloud Computing - Economies of the Cloud: Going from 350 Servers to 35

Economies of the Cloud: Going from 350 Servers to 35

Written by Eric Novikoff

Last week witnessed some interesting pronouncements at the Structure09 conference, in which the industry's leading lights had the chance to boast about their accomplishments. One that caught my eye was that LinkedIn "only" uses 350 servers.

I'm shocked to hear that a text-based service, even with millions of users, uses 350 servers, and that someone would think that was a relatively small number. Working here at ENKI with customers who do similar things to LinkedIn, I'd always assumed it would be 20-30 servers, at most. I can't help but think that LinkedIn isn't taking advantage of the latest automated virtualization management technologies, or there is some sort of inefficiency in their system. I'd love 350 servers of their business, but I think I'd be embarrassed to charge them for all those servers.

As an example, I recently participated in planning the conversion of a major email management site to our cloud. They have about 40 servers running a windows-based infrastructure. I believe, as do they, that their software implementation is relatively efficient. In our cloud, after removing clustering and failover pairs - no longer necessary due to automated hardware failure recovery that brings the app back online within the same amount of time as a cluster can fail over - they only needed 20. When we added automated workload management, which scales actual server use to demand, they averaged 4 servers over the course of a day (which is what they actually pay for.) That big reduction came because their demand was highly variable, so their servers had averaged 5% utilization over the course of the day. (See graphic at end of this blog entry.)

Overall, they saw a 10x reduction in the number of servers needed to run their site. That's a cost savings for sure, but also 10x less greenhouse gases, something to be proud of.

Applying a similar method to LinkedIn, they should be at around 35 servers, using technology similar to what we use (whether in their datacenter or ours.) What am I missing here?