Microsoft yawns at Google's chillerless data center antidote Microsoft wouldn't be surprised if Google is using some sort of custom-built mystery software that automatically shifts workloads between its mega data centers. After all, Microsoft is doing much the same thing. "We are at such an enormous scale. Think about this world where many data centers and hundreds of thousands of …
""We are at such an enormous scale. Think about this world where many data centers and hundreds of thousands of servers are running search and enterprise services and all sorts of services," Microsoft data center chief Arne Josefsberg tells The Reg."
300 million searches a day say. Works out as 3472 a second, so I hope they are not using hundreds of thousands of servers for that.... What are we saying here? 30 seconds of dedicated server time per query? Obviously not!
Methinks you are talking up your infrastructure.
""And while each one can be energy efficient, you now need two of them."
That's a strawman argument as I believe Google is now closer to 50 datacenters, not two. If you have 50 datacenters, and by eliminating chillers you now make them 50% more efficient, then by adding an additional datacenter for load shedding you are now merely 49% more efficient.
We're so great, the human race. One day there will be 2 maybe 3 companies, with hundreds of datacenters, thousands, nay millions of servers constantly....searching the other company's servers for information until all we have is three sources of info, all indexed by the other two!
Then we will really prove how great mankind realy is!
"These infrastructures that we run - and that Google does too - are so large, you can't really rely on individuals to manually make these decisions on an application failing-over from one [data center] site to another. Essentially, it all has to be built into automation software that makes these types of decisions."
Where's the skynet icon?
"What are we saying here? 30 seconds of dedicated server time per query? Obviously not!"
why not? sure it doesn't take 30 seconds to execute a search, but how do you think google get their indexes? they don't just appear on the server fully formed. Then, there is the assumption of an even distribution of searches. Surely the big peak is 9am-9pm US time, roughly. Or perhaps its Japan or Korea, but I certainly doubt its an even distribution.
"These infrastructures that we run - and that Google does too - are so large, you can't really rely on individuals to manually make these decisions on an application failing-over from one [data center] site to another. Essentially, it all has to be built into automation software that makes these types of decisions."
Were you using all that to store T-Mobile Sidekick customers' data?
Didn't think so....
but the difference is, Google's works.
Clippy: "I see you're trying to failover your mission-critical application, would you like me to:
a) transfer it to the most suitable data centre, but not tell you which one so you never find it again;
b) spend so much time trying to decide which data centre is best that a fire starts in this data centre;
c) just kill the application now, to save you time later?"
@J 3 - you said it first, but I'll add that "cloud evangelist" must be a new euphemism for idiot.
""..and enterprise services and all sorts of services," Don't be a tool."
No sir, 300 million is *Google*'s search volume (it was reported as 200million some years ago, I've added 50% ontop for growth), Bings traffic is a tiny fraction of that. I used it to get a high upper measure, and that tells me she is exaggerating way way beyond plausible.
Bing market share is say 10% of Googles, double it to allow for Enterprise whatnots (not plausible, but hey, I'm trying to give her the benefit of the doubt here), then thats 700 queries a second (350 Search, 350 Enterprise) on average.. supposedly run on 100,000 computers????? No way.
""What are we saying here? 30 seconds of dedicated server time per query? Obviously not!"...why not? sure it doesn't take 30 seconds to execute a search, but how do you think google get their indexes? they don't just appear on the server fully formed. "
Even if it was 0.2 seconds of dedicated server time per user, it would be 50x too big, because I'd anticipate 50 users per server, not *one* user per server. Even allow a doubling of time for the back crawl and index, the number is so far out as to be nonsense.
I'll tell you what, there's an easier way to solve this. Microsoft can simply state to shareholders on record how many servers they have for this search and enterprise stuff.
Will they do that? I thought not.
A digital thermometer in every data centre can report to every font-end load balancer and it's a pretty easy task to spot those where temperatures are rising and direct traffic to those which are not, even tell those data centres you're no longer directing data towards to power-down for a while.
Okay, the algorithms may be fancy to make it optimal, but I don't really see this as anything "magical" or "mysterious". Indeed, it could perhaps be simply implemented by telling the load-balancers that heat build-up is extra load and they'll automatically choose least-loaded / cooler back-end handlers.
Lightweight coolers can be installed for the cases where local ambient temperatures or total workloads rise or where there's a delay in load-balancing being effective.
Seems all very pedestrian to me. Maybe I'm not understanding what all the fuss is about.
Mine's the one with the "Acme HVAC Services" logo.
"If you're offloading data like that, essentially that would mean you would have to have a second data center with the same infrastructure somewhere else," he says. "And while each one can be energy efficient, you now need two of them. So the net is actually energy inefficient. You need a lot more infrastructure."
Will say this is actually a good argument for the efficiency crowd. Not to mention the cost of multiple data centers. Also the question being raised about the links being cut. What happens then? Does Google mysteriously load shed to the cloud? Oh wait. . .FAIL
Great concepts on both ends but in all honesty I think M$ is actually onto something here
Ahgr, what a dreadful antivocabulification.
And as for possible "obviously stupid" solutions by two of the largest data corporations in the world: however badly they may or may not have phrased it for the general public, don't you think you're quite an idiot if you argue either way in comments from your lazy chair, having none of the facs at hand? What with famous secrecy ("chocolate factory", umpa-lumpas and all that) and so forth?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
