Slashdot Mirror
How Internet Data Centers Waste Power
Rick Zeman writes "The New York Times has extensively surveyed and analyzed data center power usage and patterns. At their behest, the consulting firm McKinsey & Company analyzed energy use by data centers and found that, on average they were using only 6 percent to 12 percent of the electricity powering their servers to perform computations. The rest was essentially used to keep servers idling and ready in case of a surge in activity that could slow or crash their operations. 'Worldwide, the digital warehouses use about 30 billion watts of electricity, roughly equivalent to the output of 30 nuclear power plants.' In other words, 'A single data center can take more power than a medium-size town.' This is the price being paid to ensure everyone has instant access to every email they've ever received, or for their instant Facebook status update. Data Center providers are finding that they can't rack servers fast enough to provide for users' needs: A few companies say they are using extensively re-engineered software and cooling systems to decrease wasted power. Among them are Facebook and Google, which also have redesigned their hardware. Still, according to recent disclosures, Google's data centers consume nearly 300 million watts and Facebook's about 60 million watts. Many of these solutions are readily available, but in a risk-averse industry, most companies have been reluctant to make wholesale change, according to industry experts."
"Buy our product or we'll agitate for standards that make them mandatory." It's shit like this that annoys me mightily about the NYT.
Dog is my co-pilot.
Using VMWare or other similar technologies, you can dramatically cut the amount of the energy you need to power your servers. You can even take advantage of on-demand servers, so that if you do suddenly become busy, it'll power up more hardware to handle the load. Great for optimizing around a 9-5 workday.
>This is the price being paid to ensure everyone has instant access to every email they've ever received, or for their instant Facebook status update.
Way to trivialize users' needs.
Crikes.
--
BMO
I have no idea how the URL got mangled when Timothy moved the anchor text to a different part of the article, but here's the correct link:
http://www.nytimes.com/2012/09/23/technology/data-centers-waste-vast-amounts-of-energy-belying-industry-image.html?hpw&pagewanted=all
And since when does a tech site need to spell out "millions" and "billions"? Are we not able to grasp mega and giga?
Oh, and the little shitty comment "This is the price being paid to ensure everyone has instant access to every email they've ever received, or for their instant Facebook status update" by the submitter is totally uncalled for. That's not what it's about, but it's the reaction the NYT is looking for. Reload these comments at -1, you'll see the submitter's "First post!" comment there. What an idiot.
Shutting down free speech with violence isn't fighting fascism. It IS fascism!
Why not make a room sized server with a kajillion CPUs, host everyone in VMs, and then you can do all the power management in one place, turn off all non necessary CPUs, ramp up as needed.
I wonder if the excess servers could be left off, and during rush periods, they could be turned on via IPMI?
I'm not a lawyer, but I play one on the Internet. Blog
It's not like we needed that power anyway.
in this letter and comment.
The most ironic point: "Should we discover (as we undoubtedly would) that tens of thousands of copies of today's NYT were printed, delivered, and sold to subscribers who never read Glanz's report, do we conclude that the NYT needs a new and less-wasteful business model?"
I know of a couple of data centres in the UK which are essentially 'full' even though there is plenty of bandwidth and space. The penalty for using too much power has also increased and enforced more.
To be fair they are in densely populated areas - both commercial and residential.
so 30 Billion watts for something like 2 billion internet users. That's not to shoddy. Probably similar to the amount of power used at the client end (though that ought to be falling as people move from desktops to laptops and tablets). Global power usage is 15 Tera Watts, so data centres are about 0.2% of energy use.
So, what would do more to save the world, made data centers twice as efficient, or make transport 5% more efficient. Unfortunately if the former that's easier.
There are so many aspects left unexplored. Part of the problem is that power is also wasted on inefficient code. Bad abstractions and poor data structures. Reasons schedule pressure and untrained monkeys doing coding in PHP. It is too much focus on the ones running the data centers, part of the problem is who they are buying their software of.
This article is simply trying to make news where there isn't any. Of course only a fraction of the power consumed goes into actual computations. For starters you need to account for cooling. Roughly speaking for every watt of server power load, you nede to account for 1 watt of cooling energy. This essentially halves the potential efficiency. In addition to that, you need to account for the amount of power it takes just to maintain state when you talking about a data center of that scale. Non-volitle memory requires and consumes power just to retain its current values. Unline Facebook and Google, most datacenters do not have 100% control over the hardware and software being run. Additionally datacenters often charge for power, space, etc and the client simply pays for what they use. In many instances efficiency is not for the datacenter to determine and one could argue that it may not even be in the datacenter's financial interest. Great strides have been made in scaling power consumption to fit computational demand but this is more of a hardware/software issue than a datacenter issue.
This is the reason there is a race for high performance chips that draw little power. Your tablet may sip power and take a few seconds to render a Facebook page, but the server sending milliions of pages needs to sip little power too. Whoever makes the best server chips wins.
The truth shall set you free!
spying isn't wasting power
A server is a sort of bulked-up desktop computer, minus a screen and keyboard, that contains chips to process data.
Schroedinger's Brexit: The UK is both in and out of the EU at the same time!
first helium balloons, now data centers. its nice that slashdot is shaming humans today on their wasteful energy practices.
i had decided that i was going to just shoot myself to eliminate my wasteful means, but i guess that would create "waste" as well. people would report the smell, the police would then arrive, news reports, etc.
maybe ill pre-dig a grave and bury myself alive. that should do the trick.
How much power is being wasted by sites that do not honor "do not track"?
Contribute to civilization: ari.aynrand.org/donate
This is lovely. Let's worry about problems that don't exist, as if we don't have enough catastrophes to worry about.
Power is money. As long as there is a somewhat unhampered economy in the locus of data centers (and there is), then every entrepreneur will attempt to economize power usage. You don't have to worry about it because the entrepreneurs that use power efficiently will eat the lunch of those that do not, ceteris paribus (all other things equal).
Ipso facto this problem will solve itself. Case closed.
In fact, now that I speculate on the possible reasons for publicity like this to be drummed up, it is to campaign for government regulations that will instruct entrepreneurs how they 'must' handle such a problem. Unfortunately nobody can write such regulations because they cannot foresee every circumstance and possibility, much less predict the future. Nobody on this earth can even tell a single other person what ideal type and amount of preparation is for power efficiency considerations. This is why we have economic calculation.
If such regulations are enacted, ipso facto they will cause the problem itself.
Liberty.
Lovely wriggle phrase: "a medium-size town".
Most people will substitute "city" or "urban" as they read, from the distinction "town and country". In some places 'city' and 'town' are synonymous, in others a 'town' is too small to be called a city, what other first-world countries would call a 'village'.
Well, it's the NYTimes, so let's try a NY definition.
http://en.wikipedia.org/wiki/Administrative_divisions_of_New_York#Town
" The town of Hempstead (Nassau County) has about 756,000 people (2000 census), making it more populous than any city in the state except New York City. Red House (Cattaraugus County), the least populous, has 38 permanent residents (2000 census)."
It's a hideously unspecific term for that hides more than it reveals, perfect for yellow journalism. And we're supposed to take /anything/ from this NYT article seriously?
First post? ...on my own submission? LOL.
No no; you see you made the mistake of shutting down your server between the post and when it was put up by slashdot to save energy. You are just too slow to achieve internet time like Google or Facebook and will never make it in the market. Let this be a lesson.
=~ s,(.*),<sarcasm>$1</sarcasm>,g if any_point_you_wish();
Take the case of me and Google. My share of their power is about 1W electric (that's usually about 3w thermal).
However, I estimate that their maps and local business info features alone easily save me at least a couple of hundred miles per year of driving. That would be about 10 gallons of gasoline per year, which is 38 W thermal that I'm not burning thanks to the info they're providing. Google provides at least a 10 to 1 payback in energy savings just for this one case.
These are completely meaningless numbers. What matters is power per user compared to the power per user spent by alternative methods or other data centers. If the data center is using one tenth the power for running my database that I would by running my own server, then the data center is a boon energy-use-wise - even if the data center could theoretically decrease their power usage even further.
In any case identifying wasteful use of energy in this way is entirely the wrong way to go about it - does the NYT plan on identifying and writing about every little way someone somewhere could decrease energy usage? If the price of energy isn't enough incentive for data centers to use less energy, then either the price of energy is too low or your notion of the value of energy is inflated or you just need to wait a bit while the energy inefficient data centers go out of business. In the first case, the solution is to include the externalities of energy usage into the price of energy through taxes (rather than the current subsidies). In the second case, well, you need to fix your views. In neither case is it relevant how data center operators choose to do their business - that is merely a symptom of the underlying forces.
This is why altruistic energy saving is pointless - if your decrease in consumption does not result in a decrease in dirty energy production, then you have done precisely nothing to help the environment. All you have done is decrease the price of energy by enough so that someone else is willing to use the energy that you didn't use at that price. Conservation does not save energy, it only saves you money and decreases the price of energy for other people. The latter point might actually be a pretty nice thing to do - people in the third world certainly have a pressing need for cheaper energy. It just doesn't help the environment.
For conservation to make any difference to net energy use you need to conserve so much that oil can be shooting out of the ground yet no one wants it because they don't have any use for more energy. Never going to happen! The only solutions are to increase the price of energy to the point where people won't want to use that oil shooting out of the ground because of the tax expense (cap and trade does this, banning use of oil does too) or to come up with another way of producing energy that isn't dirty and is cheaper than oil. The first solution isn't going to make you any friends as a politician and it only works if almost everyone in the world is in on it (never going to happen!). The second solution is the only hope if you want to avoid more CO2 release. Though I guess at the point that we are all drowning in rising oceans perhaps there will be enough will to really make the first solution work. I doubt it, though - a third alternative is geoengineering, which will be necessary anyway at that point. Feel like rolling those dice?
I'm part of the problem. Wish I wasn't, but I don't seem to have any choice.
I run a small web site, and if it goes down, there are various consequences in my personal and professional life that can be extremely annoying and embarrassing. To stay sane, I need the site to have good uptime. Over the years, this has caused me to gradually migrate to more and more expensive webhosting, now ~$100/mo.
The average load on my dedicated server is extremely low, so it's basically like one of the extremely wasteful boxes described in TFA. My site is basically I/O-intensive: I serve big PDF files. In terms of CPU, I'm sure the site would run fine on a low-end ARM, or as one of a dozen sites running off of the same Celeron chip. So by comparison with either of those hypothetical, energy-efficient setups, virtually all of the electrical power is being wasted. I'm a small fry, but there are millions of sites like mine, so I'm sure it adds up. (It would be interesting to know how much of total server-center power consumption comes from the "long tails" of the distribution such as Google and Facebook, and what percentage from cottage industries like me.)
There are basically two problems. (1) Nobody will sell me high-reliability webhosting on low-end hardware. The only way to get energy-efficient hardware is to get cheap webhosting. I've tried cheap webhosting. Cheap webhosts have low reliability and nonexistent customer service. (2) Sometimes you get spikes in demand, and you want some excess capacity to be able to handle it without crashing the server. Maybe you get slashdotted. Actually, in my case one thing that has been a problem is that some people apparently run IE plugins that are supposed to accelerate large downloads, by opening multiple connections with the server. When these people hit my server and download a large PDF, the effect is very much like a DOS attack. My logs show one IP address using 300 Mb of throughput to download a 3 Mb PDF. I've written scripts that lock these bozos out ASAP, but on a low-end machine, these events would bring my server to its knees instantly.
Find free books.
Figure out how to 1) make computers boot faster and 2) reliably support suspend/sleep/etc. Using free software, of course, and please don't restrict yourself to linux.
Most servers have crap boot times, fix this plx.
Tablets are (apparently) ousting desktop PCs and laptops as consumer devices. These are by necessity low-consumption, hence low-capacity devices (as in, they can barely play an HD video without screeching to a bloody halt), they're certainly not going to be doing any what an 80's admin would have considered big iron work. This would be left to... well, big iron. The infrastructure is already there; thin clients, virtualisation on multicore beasts that can chew through 4k CGI rendition in practically real time, cloud storage and fast broadband. Hey, did I mention the word "theoretically"?
There's your justification. Thirty Gigawatts is what 66 million average desktop computers (at 450W a pop, not including displays) consume. Think about what 66 million netbooks, or tablets, consume?
Asus EeePC (Atom): 40W (according to the wallwart I've got plugged into the side of mine) each. 2.64GW.
iPad: from what I've read, 10W a piece - and the screen uses 6 of that. 660MW.
Sod it, add them together. That's 132 million computers, accessing a virtualisation service for a total power cost of change from 33.5GW.
That's a might less than 132 million desktop machines doing their own thing yet costing 59.4GW *on their own* - not including the aforementioned server infrastructure. Why aren't we doing this?? According to Gartner the number of personal computers in use around the world hit one billion way back in 2008. This is slightly more than significant.
Commence to shootdown citing personal security concerns in 5... 4... 3... 2...
Operation Guillotine is in effect.
"Oh, sorry, can't service your request, we're at capacity now, come back later when you've totally lost interest or forgotten about this."
How about this: "We've gone green, and we keep some of our servers turned off. We'll have the page ready for you once you're done watching this interstitial ad."
Part of the problem is that power is also wasted on inefficient code. [...] Reasons schedule pressure and untrained monkeys doing coding in PHP.
At work we have a reasonably trained monkey coding in PHP. What language would you recommend that is more efficient for a web application, balancing programmer efficiency with runtime efficiency?
As long as there is a somewhat unhampered economy in the locus of data centers (and there is), then every entrepreneur will attempt to economize power usage.
There is not an "unhampered" energy economy. The energy economy is massively subsidized in several ways. The US government has spent trillions in wars and foreign aid to secure energy supplying areas. Our natural gas glut right now is going to be paid for by future generations in the form of devastating environmental damage, like damage to our water tables. Our continued use of fossil fuels in general will also be paid for mostly by future generations in the form of the costs of global climate change. If the full cost of energy was represented in the price of electricity then data centers would behave very differently.
All things are interconnected. You claim that data centers are a market that behaves rationally, but data center's main feedstock comes from a market that is distorted to the point where it is threatening the continuation of our civilization.
-- QED
I know the term is horribly abused on a regular basis, but that's the whole point of cloud computing. It looks like a horrible idea because today it usually is, but eventually we'll get used to it and figure out how to make it work most of the time and then we can have a lot less idle resources as they can just be turned off. Even if it doesn't get you entirely out of colocated resources, if it can decrease the amount of hardware you have to have lying around doing nothing most of the time there's a place for it.
"You're right," Fisheye says. "I should have set it on 'whip' or 'chop.'"
Slashdotters are right - it must be ok to waste power because: 1) It's less than other power we are consuming (or wasting?) 2) it's in the name of progress 3) It's not us who are wasting it This is the same thinking that says that if you have a million dollars, it's ok to just thow $100k in the garbage can. Thank G*d slash-dotters don't run the world. Stick to writing code and keep your opinions out of saving the planet, because there wouldn't be one for you to code in if we listened too closely to your my-math=your-reality reasoning.
Sent from my ENIAC
This is one major reason that companies (even very large companies with "money to spare") are moving towards virtualization with incredible speed.
I'm not going to go digging for numbers right now, but the statistics show that something like 100 percent of Fortune 100 companies use virtualization, and perhaps 85-90% of Fortune 500 companies.
The larger virtualization solutions will actually take the servers that are idle, migrate them to another host machine, and power down/suspend the "extra" machine(s) that was/were being used during their core business hours.
Virtualization also allows for spikes in cpu/network, and then can take that power back when everyone goes home (a print server, an intranet web server, a domain controller, etc). So, physical machines actually DO get turned off when they aren't being taxed, and with more and more "software defined networking" the interconnects between systems can be scaled and moved also.
Now, I don't know how the big players are using this (e.g. Amazon, VMware, Rackspace, Google). I can't see inside their datacenters, but one would think that something like AWS would have a huge stake in saving power by turning off idle instances and moving VMs. Not only for the power savings from the server directly, but for the (approx) 30-40 percent more energy that it takes to cool the physical machines.
It's also worth noting that larger companies are putting their datacenters in areas with plentiful (cheap) power. Places like Washington state, with hydroelectric power and a cooler average ambient temperature, allow for a huge savings on power right off the bat. Add things like dynamic scaling of server and network hardware, lights-out datacenters, and better designed cooling systems (look at Microsoft's ideas), and there is a huge power savings across the board.
How much energy does the NYT use to print paper copies of the newspaper, distribute and deliver them, harvest the trees and process the paper? Now compare that with the energy that the online NYT uses. Which allows for more people to view the publication for less energy? I'm positive that it is the electronic version.
Grandpa: My Homer is not a communist. He may be a liar, a pig, an idiot, a communist, but he is not a porn star.
This is the same "problem" that faces airline companies, taxi drivers, power companies, cell network operators. Consumers pay for these services by usage and so total revenue is proportional to average use but the costs are heavily skewed towards capital costs and so are proportional to the peak load that you can service. In that case, there's a fundamental tradeoff -- either we have to degrade service when demand hits the 95th percentile (just as an example) or we have to figure out a way to pay for the extra capital investment that's not needed 95% of the time.
There's a few alternatives you can do:
(1) Overprovision and soak it up into the price structure for all consumers. This is what most power companies do -- they build enough power generating capacity for peak load and then charge a bit more per KWH to make up for the increased outlay.
(2) Overprovision and charge extra at peak. This is the airline solution -- they always have service available but under contention the last few seats are exorbitantly expensive. Essentially those that need peak service are paying to leave a few seats open all the time in case they need them.
(3) Don't overprovision: this is the taxi solution. This means that service degrades significantly under peak demand -- anyone trying to get a cab home on a Saturday night in a major city has experienced this. Those that do get a cab pay the usual fare, everyone else waits around a while. This is also the solution that California has routinely deployed for their inability to provide peak power during heat spells -- same price for everyone but rolling blackouts for the unlucky few.
That's it -- there aren't any clean answers when you are making compromises between peak availability and average efficiency. You've either got to pay for the extra capacity when you don't need it or else you have to suffer when you don't have the capacity when you do need it.
Think of how many data centers have dedicated appliances for filtering spam. If they want to save on power they should take some actual action against spam instead of just being reactionary.
The data centers (and to a larger extent ISPs) remind us that spam is an economic problem. It is costing everyone money every day, so that a handful of spammers can make a lot of money pushing fake pills, fake watches, etc. If the data centers seriously want to reduce wasted power they should instead invest some human time and effort into making it more difficult for the spammers to make money. Do that, and everyone wins. Keep filtering spam, and nobody wins.
Damn_registrars has no butt-hole. Damn_registrars has no use for a butt-hole.
I'm in the DC industry, we pay for 100% of the power we use, so we only use what we need. True we have to keep the lights on, the servers on, the cooling, and the rest of the support equipment. There are efficiencies lost in power conversion and everything else. We cannot control the idling servers -- that's what virtualization does, it helps move the loads to a common machine and eliminates waste. I get what the article is saying, but a DC normally can't control the customers processing.
So there.
My first thought when I saw the post was: "bean counters strike again: Look at the wasted resources that aren't used all the time!"
having extra capacity to handle peak usage is not waste. Never will be. Making sure the extra capacity is efficient is another question, maybe the article linked talks about that, but I always get riled up when someone sees idle capacity and calls it waste.
AB HOC POSSUM VIDERE DOMUM TUUM
OK, so they're wasting power. What is it that the NY Times wants me to do? Stop using these companies' services? Get my green on, and start complaining about the effect on the planet?
All these websites are free--I'm not paying for what they waste. I'm not investing in them. If they want to waste power, that's their poor cost-management. It seems like a no-brainer to look at where money is being wasted and try to fix that.
We're not short on electricity and having power outages all the time. And again, they're paying for the power. What should we do if they were twice as efficient? Shut down 15 nuclear plants?
Really though, I would like to see more server-side processing and a lot less flash. I'm tired of websites (news sites being the worst offenders) that have 20 flash ads, especially the ones that tie up way too much CPU time for something like a button...that opens a link. Or a mouseover. No idea how many useless conditional statements and random loops are going on there.
Also, NY Times is writing about how bad data centers are, when a huge fraction their business is dependent on these data centers being always available so people can get breaking news from them.
The normal way of allocating resources is simply by cost/ benefit. When the cost of powering idle servers is greater than the loss of customers due to latency problems, providers will cut back. Why make things more complex then that. Incidentally the electric power providers have far worse problems building capacity for peak loads.
Perhaps those of you who fear nuclear power should advocate washing clothes at night. That might reduce peak loads more that data center schedule changes.
A minor typo in PHP and you won't get so much as a warning
Even if you set an error handler that converts all errors to ErrorException?
Virtual machines - working around the limitations of operating systems that cannot effectively do a few tasks at once.
There is also the security aspect of having Chinese walls between things on the same system, but it was initially as accidental as NAT stopping people getting in, and despite a lot of work to avoid exploits it's not really any better than some of the solutions within an OS.
so, that would be like a million servers, maxing out their 300 watt power supplies for every second of every day? Hrm, sounds a bit unlikely. Well, okay, figure it consumes about just as much power to cool as the server consumes, since it is (in theory) not putting any chemicals into high chemical potential...well, yes, then there's the inefficiency of air conditioning, offset by the natural cooling by fans and heat sinks (of buildings, not servers). Still that's a lot of power. I know that cisco gear has no concern for one's electrical bill and that the old Sun servers did live up to their name in some ways by pumping out heat...but still....
You know, there's this whole country Iceland with lots of cheap electric power and natural cooling, that doesn't consume any fossil fuels to produce it. Maybe we could put some servers there?
1. Implement orchestrated Virtualization and Paravirtualization where security, regulatory and other compliance requirements allow... but only duplicate kernels and guest operating systems where it makes sense. (i.e. Eucalyptus, OpenStack, Xen, KVM, LXC).
2. Persist data separately from compute resources in order to reduce idle CPU and disk. (i.e. S3, Walrus, Swift, GlusterFS, ZFS, Ceph).
3. Make hardware improvements... perhaps using solid state drives (where higher I/O is demanded) or using un-cabinetted server grid technologies such as OpenCompute to improve air flow and reduce cooling costs.
4. Provide self-service dashboards and APIs for elastic provisioning of resources that can be monitored by IT service managers or resource brokers as well as the user who is consuming these resources and offer incentives for intelligent resource consumption.
- Asher Bond
Elastic Provisioner, Inc.
The NYT's whole business model is based on exposing dirty little secrets that sell column inches of newsprint and ad space.
If they had been reading IT/Web technology papers or /. articles they would have seen the ones regarding Facebook and the open source & open hardware models being developed. Not to mention the wildly inventive things Google has done over the years, which we only find out about when they have already moved onto the next phase.
Others have said it regarding the basic method of calculating service level versus service maximums yet the NYT still doesn't get it. Of course, buying the digital version of their product is more expensive then the print. Yep, they got all that technology stuff figured the heck out. 'Cause that's what they do there in NYC. Figure things out. And sell ads.
I've talked in the past on why liquid fluoride thorium reactor (LFTR) should be the immediate future of nuclear power generation.
Here's one thing I haven't mentioned: LFTR's can be scaled down to 50-80 MW powerplants which are amazingly small and require very little real estate to operate one. Because of its very small size, an 80 MW LFTR could be almost near the site of the big server farm itself, and that could mean the server farm doesn't need a land-wasting big solar power farm nearby or have to be located in an area with a lot of power generating capacity (what few people know is that Apple located in server farm in North Carolina because there are plentiful coal-fired power plants generating a lot of power in that state, which means Apple can continue to operate the server farm in case there are problems with the solar power farm).
Has anyone actually ever seen an overheated Ethernet socket?
Isn't Apple's server farm in NC going to be mostly solar and all green? Certainly a good precident.