Facebook’s Open Compute could make traditional servers obsolete, says Peter Judge
The Facebook-backed Open Compute initiative is planning to blow servers apart. That’s not hype, it’s what the no-frills initiative for data centre computing is actually trying to do. Last week we saw some indications that it might succeed.
I’ve described Open Compute as “maximising LOLs per Watt” elsewhere and that’s a fair description. Facebook and other cloud providers see massive demand for their services, mostly to deliver consumer-oriented stuff, like family photos on Facebook. To carry on delivering that, Facebook and its competitors have to spend as little money on hardware as possible, and burn as little energy as possible – in other words, to extract the maximum amount of social interaction (LOLs – of course…) from each Watt of power that goes into their data centres.
Server makers are collateral damage
Facebook – and other members of Open Compute – have no other agenda than that. So it’s just a fact of life that some of the industry’s biggest players will be pushed to one side by the initiative, which could reshape a large part of the industry.
Because in Open Compute, some of the largest buyers of servers are dictating terms to the server industry. They are saying they don’t want compete servers in one box, but would like to have separate power supplies aggregated on one shelf.
They don’t want to have motherboards that are tied to one silicon vendor’s products – the Group Hug board has a new slot design that allows them to swap chips.
A lot of this isn’t new. Any big web service provider has the muscle to get its own way in the servers it uses. Google, Apple and Amazon have all enforced their own tweaks on hardware makers, or bypassed the big names and gone straight to original equipment makers based in China or Taiwan to have servers made to order.
The difference with Open Compute is that the specifications are being shared. Google has always seen its server designs as business critical, never to be opened up to rivals.
By putting its specifications into the Open Compute project, Facebook has done something different. It has encouraged input from others such as Intel.
And Open Compute has allowed the next tier down to benefit from open hardware: Rackspace has adopted the Open Compute designs as the basis for its own systems. And there’s no reason why Open Compute designs should not be offered to much smaller companies.
All of which means trouble for the big hardware makers. Adopting Open Compute ‘s no-frills designs would cut their margins. Ignoring them would limit their markets.
It could be a major change for the industry – and all as a result of maximising LOLs per Watt.
What do you know about open source stuff? Try our quiz.