>> ^joedirt:
"because UPS is an integral part of a server."
Yeah, sure. And my spare tire pressure is a critical statistic I brag about my hotrod.
Duh, the UPS has nothing to do with server operation nor power efficiency. Look at the spare tire on that Posche!!!! Wow, that's hot.


When you get a flat tire on your "Posche", it only inconveniences you (and your passengers, if any). I'd say that's a maximum of 4 people being inconvenienced, maybe 6 if you have one of those silly "Posche" SUVs.
Now, take a look at google - a single data center has capacity for 45,000 servers, and each server probably handles 10s to 100s of simultaneous requests (being conservative). Now lets say the data center loses power, without the UPS, every request is lost.
30 seconds later the generators kick in, all the machines reboot in a few minutes, and facility is back online, but in the meantime you've just inconvenienced a million people. Not really good for PR....

So for a business like Google, a UPS IS an integral part of a server.

As far as efficiency - take a 10 megawatt data center, assume you have a central UPS (for reasons mentioned above) that is 95% efficient. That means you lose 500kW of power in the form of heat. If you replace that UPS system (as a whole) with one that is 99.9% efficient, you lose only 10kW of power to heat.

To put this in perspective, the average household in America consumes about 14,000 kilowatt-hours (kWh) per year, according to the Department of Energy.
That's an average of 1.6kW per hour. So by going from 95% eff to 99.9% eff UPS system, google saved 306 households worth of electricity.

The main reason for the effeciency gain is that you aren't going AC-AC-DC-AC-AC-DC, like a traditional server room, but rather AC-AC-AC-DC.
(AC = alternating current, like power lines ; DC = direct current, like batteries; hyphen represents a conversion, none of which are 100% efficient)

Joedirt, at first after reading your comment I felt like discarding this (I hadn't read the CNet article before submitting, only the short Engadget take on it, and I was never really interested in servers, so admittedly, I didn't fully understand the specifics of data center efficiency), because I don't like to spread false facts around.

But now that I've wised up - sorry, I'm keeping my video and for now I'm also keeping my (admittedly sensationalist) title (but that's because it's exactly what Engadget reported, and I can't think of a short enough alternative that'd include the UPS but I'm open to suggestions).

Why? Simply because UPS is an integral part of a server. Sure, it's not the server as a whole that 'boasts 99.9 percent efficiency', just a component of it, and yeah, at first the title might imply their PUE is ~1.01, but the video description sorts it out. And besides, from what I read some of their PUEs (as low as 1.12) are quite phenomenal as well. If you suggest a more accurate, appropriate title, I'll gladly change it, though

P.S. All in all, thanks for your comment; honestly, it's always good to have someone that can provide additional, accurate info/take on what's being sifted.

>> ^quantumushroom:
9) The majority of the Antarctic is 8 degrees below “normal” (again, whatever that is)


"Satellite imagery from the National Snow and Ice Data Center at the University of Colorado at Boulder reveals that a 13,680 square kilometer (5,282 square mile) ice shelf has begun to collapse because of rapid climate change in a fast-warming region of Antarctica.

National Snow and Ice Data Center, March 25, 2008

And....

"Ice growth slows; Arctic still warmer than usual"

National Snow and Ice Data Center, December 8, 2008

Net neutrality is more about extending the QOS Trust boundary out to the users home to apply metrics to certain data streams to either throttle or block entirely selected applications based on any policy criteria that the ISP chooses.

QOS boundaries are technical abstractions, they basically show where devices or data-streams trust (in QOS priority flags) can be kept, and passed on (ergo, allowing QOS policies to be applied to certain traffic).

At present these boundaries are inside the ISP's own cloud (generally the distribution network, ala the routers/switches at your local peering point / data center for your ISP - PAST your local telephone exchange).

This makes it a nigh impossible task to apply QOS to individual applications and data-streams without incuring SIGNIFICANT performance decreases on the distribution network. Its much more efficient to apply that kind of stuff as close to the user as possible, and for that very reason, the ISP's have been trying to push the trust boundary out to the exchange so that they can throttle applications / services that they deem worthy of it, so that they dont have to spend as much money on upgrading distribution / core bandwidth when user's speed/density gos up with the technology.

The part about broadband penetration is also a very important one. Basically hes saying that the govt. and the FCC have dropped the ball, allowed larger corporations to eat up all the smaller ones (either through predatory pricing schemes or straight-forward buyouts) which in turn destroys competition in the market.

Competition is the ONLY force that drives innovation and efficiency, resulting in BETTER services, to MORE people, at LOWER costs.

Without that, none of those 3 fundamentals which push an industry forward, ever see the light of day...which is basically why you've seen broadband penetration decay over the past 8 years.

We have been having a similar battle in Australia with the incumbent ISP that owns the entire phone network to every single house....basically they have conflicted interests between their retail department and their wholesale department, allowing them to use their enormous ammounts of economic AND political muscle to strangle the other ISP's out of the arena, while at the same time holding back advancement in network upgrades.

15 years ago they had a plan to run out FTTH to everyone in the country, when it was still under control of the govt. When that was privatised, they shelved those plans and have been artifically capping speeds, and styming investment ever since and it has been a disaster on our digital economy...but the shareholders are all happy, so who cares, right ?

This is the crux of the argument...without regulation, you allow the incumbents to strangle the industry at the expense of the consumer choice, all so their share-prices can go up at monopolistic rates.

Its a bad bad thing that must be stopped from occuring at all costs.

Im a network engineer by trade, and currently studying telecoms engineering and hold a comp sci degree, this is my bread-and-butter

>> ^MINK:
^call me in 2020 and laugh at me if i was wrong lol

Sorry, can't. Verizon won't let sip packets to your phone number through the data center in Pittsburgh, unless you remember to pay your "let people in Pittsburgh call me" fee to my local net gatekeeper(as well as yours, and everyone in between).

The site seemed slow as molasses yesterday to me. All day I was having mini withdrawals becuase a page would take 20 seconds to load.

and looris, the costs for bandwidth and servers varies depending on your deal. If you are lucky enough to know someone who works for and ISP or at a data center you can get this stuff fairly cheap. If not im sure its easily around a couple hundred a month for server and bandwidth.