Google Joins Facebook's Open Compute Project

judgecorp writes: Google has elected to open up some of its data center designs, which it has -- until now -- kept to itself. Google has joined the Open Compute Project, which was set up by Facebook to share low-cost, no-frills data center hardware specifications. Google will donate a specification for a rack that it designed for its own data centers. Google's first contribution will be "a new rack specification that includes 48V power distribution and a new form factor to allow OCP racks to fit into our data centers," the company said. "We kicked off the development of 48V rack power distribution in 2010, as we found it was at least 30 percent more energy-efficient and more cost-effective in supporting these higher-performance systems." The company said it hopes to help others "adopt this next generation power architecture, and realize the same power efficiency and cost benefits as Google." Google hasn't submitted a proposed specification to the OCP yet, but the company is working with Facebook to get that done.

weasel words

[Gravis Zero]

Google's first contribution will be "a new rack specification that includes 48V power distribution and a new form factor to allow OCP racks to fit into our data centers,"

they aren't going to "allow OCP racks to fit into [their] data centers," they are trying to make OCP adopt their design.

just sayin'

Re:

[Frosty Piss]

they aren't going to "allow OCP racks to fit into [their] data centers," they are trying to make OCP adopt their design.

Well, boo hoo. Two companies with some of the largest and most efficient datacenter programs on the planet think they know the Secret Sauce? Surprise, maybe they do. But just go ahead and ignore their obvious expertise because you don't like "Just Be Evil" Google and, well, "Just Be Evil" Facebook...

Re:

[Junta]

The problem is that Open Compute represents some wasted potential. When Open goes well, there is a large chunk of consensus around which good collaboration is done. For Open Compute, no two big players have really agreed on anything significant that wasn't otherwise a settled matter in the wider industry. The connector of a 'daughter' card is agreed upon, but there are at least 4 different conflicting mechanical specifications for what attaches to the connector. This is probably the closest to consensus

Power supplies?

[glitch!]

Okay, I am interested in 48v (nominal) feeds. Where can I buy a decently priced PC power supply? If Google is buying them left and right, they should be easily available and cheap, right?

Re:

[Jack Griffin]

I seem to recall reading something years ago about designing a (mostly) DC powered house, and 48V was the number they landed on too.
Since most things in your house are DC anyway (I have gas cooking and heating), and with most only needing circuits in the region of 10 metres or less, I can't remember why this didn't become a thing?
Surely it would make it a lot easier to switch to off-grid battery storage solutions if everything is DC?

Re:

[4wdloop]

With LED lighting it could be more possible. However, a microwave, hair dryer or iron would draw a ~20A current (~1kW power). And this requires a thick wires. Gone are electric heaters, ovens and AC etc. Smallest household hookup is 2x100A @ 110V about 15kW peak power which equivalent 48V DC would require >300A, gauge 0000, 0.46" wire in diameter.

Re:

[GuB-42]

For power transmission, a large voltage is often a good thing as it reduces losses and allows for thinner wires.
We also know that DC starts being dangerous at around 50V. It means that the ideal is just less than 50V : relatively safe and can power devices in the kW range without unreasonably thick wires and risks of overheating.
48V fits well and as a bonus, it is the nominal voltage of 4 car batteries in series.

Telco has been here before

[satsuke]

Telecom equipment has been using these 48vdc power systems for decades.

That industry certainly hasn't had the same goals in mind as far as efficiency or getting the kind of density that Google et al would need, but the designs should have a lot in common regarding physical plant designs.

e.g. in the telco space, N+1 (need +1) or better is pretty standard, with the distributed nature of Google and other large compute clusters being (likely) more tolerant of a given node or blade server failing without impact on the whole.

I still feel uneasy about the name

[U2xhc2hkb3QgU3Vja3M]

Is anyone else a little scared that OCP is now something real and that it's computer-related on top of that?

Re:

[esperto]

As long as they don't call their security model ED209 (with a built in 20 seconds to comply), I don't think we should worry.