I have read this article when it first appeared on Hacker News: http://news.ycombinator.com/item?id=542531 and I was in doubt whether to believe it or write it all off as April's Fool joke. The design looks interesting and reasonable, but... who releases serious information on April,1 really?! If they were actually serious, I was sure they would release this a day later or earlier.
Now, quite some time later, I read this again and I am not convinced by the article. The comments also do not offer evidence that this is not a joke. Nor the video that you posted, until the second part. In the first part all it shows are some containers and industrial machinery; nothing indicates that there are really servers in those containers or all of this stuff is actually Google's. Now after the second part I believe that it all was for real, after all.
I wonder if Google is playing with us on purpose, releasing the information on April,1. Or Maybe they just don't care a single bit about such things?
Fascinating! That's an off-the-shelf SLA (sealed lead acid) battery like many UPS/alarm systems use, which need replacement at least every 3 years (or much sooner if they are deeply discharged).
But I would happily pay $2, even $5 more for that feature on a motherboard.
From the article, the motherboard only requires one 12V input only, instead of the usual array of voltages (+-5, +-12, +3.3, etc). Having charging circuitry to keep the battery operational (trickle charge) and being able to provide power to the motherboard when mains power fails is mostly straightforward and cheaper in this case. Not much different from what you would find on a regular UPS, but removes a few conversion steps - (DC 12V) ->(110/220V AC)-> DC - improving efficiency and reliability.
Yep. This is old news. But more recently other companies have also decided to implement on-board batteries on their custom servers, most notably Facebook:
And yes, this story was real, despite the April 1 date. I attended the event, and saw the server. It definitely seemed odd to schedule a technology unveiling on April 1, but Google apparently launched Gmail on April 1 as well.
So given the amount of research that Google pump into getting the best components for the job, are we to take it that all their testing has concluded Hitachi hard disks are better than the alternatives?
It might also be that their cost for Hitachi drives is enough to offset any higher failure rate they may have. Could go either way on that front, really.
"Hitachi manufacturers the safest and most reliable hard drives, according to the Storelab study. Of the more than 200 Hitachi hard drives received, not a single one had failed due to manufacturing or design errors."
// WD and especially Seagate (Barracuda 7200.11 had major firmware problems) failed miserably. 500Gb or bigger drives were the worst for all manufacturers. Samsung was almost as good as Hitachi.
"Storelab notes that read/write head failure is somewhat characteristic for WD drives. Failures primarily occur as a result of physical impact or overheating (WD heads can be sensitive at temperatures above 45°C).
Western Digital's construction makes drives particularly vulnerable to shocks and pressure. Unlike other manufacturers, WD does not secure the hard drive axle with a separate screw to the drive cover. "
"Google has patents on the built-in battery design, "but I think we'd be willing to license them to vendors," Hoelzle said."
Can you patent a device you use as part of your business but don't sell? I don't think you could get away with calling it a business practice, and I don't think you could get away with protecting a patent on a device you never try to market.
> Can you patent a device you use as part of your business but don't sell?
Yes; I can't think of any familiar examples offhand, but it happens a lot. As a baby lawyer I wrote a patent application for an automatic pizza-making machine, invented by the owner of a mom-and-pop restaurant, who IIRC never marketed the machine.
> I don't think you could get away with protecting a patent on a device you never try to market.
In the U.S. (and most other industrialized countries AFAIK), the patent laws don't require a patent owner to market, or even try to market, a patented invention. If the patent owner was just trying to keep others out so that no one was using the invention, a court would probably take that into account in determining (i) whether to grant a injunction against further infringement by the defendant and (ii) what a reasonable royalty would be for the defendant's infringement. But that wouldn't affect the validity of the patent itself.
Yeah, totally irrelevant. Today's servers are powered by pixie dust and rainbows instead of batteries and circuit boards. I learned nothing from this article!
I learned from this article the first time it was posted; I have learned nothing further from the following N reposts. If we allow "new to you" articles on the front page it will be filled with old stuff.
A certain amount of churn is required to continually inform new users. Yes, too much causes the older users to leave from boredom. To little, however, and the newer users cannot gain an appropriate grasp of the common knowledge expected, so references and assumed points will pass them by in conversation.
The other option is the one humans generally end up with. Knowledge barriers between generations. There would need to be different versions of the site every year or two so that new users could go through their learning period uninterrupted, and older users could be secluded to their advanced topics. This makes it harder for older users to educate newer users, and leads to rehashing of debates once closed. The older users will peter out over time, while the newer will waste time spinning wheels they could easily have been advised were explored and pointless.
Complaints are only needed if there is an excess of redundant articles. The occasional rediscussion simply keeps ideas alive and flowing.
In 1992 we built PCs for a low noise astronomy camera with the PSU replaced by a 12V car battery to avoid the noise of a switchmode PSU