Slashdot Mirror
Folding@Home Client's Performance Impact Measured
EconolineCrush writes "Trying to convince your boss to let you run Stanford's Folding@Home client on the machines at work? Here's an article that measures the performance impact of running the Folding@Home client that might help. The article examines the client's impact on the performance of business applications, games, workstation applications, and more. When set up correctly, the Folding@Home client can be run transparently in the background with only a negligible impact on system performance, which means your boss has one less reason to turn you down."
this article isn't going to convince him.
You would be running the application on THIER machines, NOT yours. Ypu boss would have EVERY right to say NO.
"It's not like your minds are as open as the source you love..." - Me to the majority of Slashdot.
United Devices is another company that does grid computing for cancer research. Which leads me to ask the following question (this may be a stupid question, but I'm bio illiterate)..
Wouldn't protein folding have some sort of similarity in finding a cure for cancer?
"The ones who dont do anything are always the ones who try to pull you down" -- Henry Rollins
...my completed results could be sent.
I ran their Linux client on a couple machines and it ran ok, didn't impact things too badly (remember "nice"?). But when it went to upload the finished results, it could never connect to the server that takes the finished data.
After two weeks of that, I pulled the client down. No one bothered to respond to my email, one person pointed to a discussion group for assistance, but since I'm already being overly generous with my time, it was more bother than it was worth.
Whats that? Oh, no, it doesn't have anything to do with work.
Yes, it will have an adverse effect on the network performance but this web site I read claims it won't be all that bad.
Oh really? I should get the fuck back to work and quit fucking around with bullshit worthless personal stuff on company time or you'll fire the shit out of me?
Yes, thank you sir, back to work sir.
Cunning linguists
Don't get me wrong, I think distributed computing projects are great, especially those focused on legitimately useful research. However, running a distributed computing client on a machine at work will likely cause it to consume more electricity. A Pentium 4 has a maximum power consumption in the range of 65W, no? So every computer you install this on is like leaving a 60W lightbulb on 24-7, year-round. If you do this with many computers, I think that may add up to a nontrivial expense that you're essentially stealing from the company, no? Just playing devil's advocate...
>> this article isn't going to convince him.
probually not, but it might be able to confuse him enough so you can convince him of something else
I mean look at all those numbers and graphs!
No! .. slows down the pr0n downloads =)
/wave ;)
Its pretty common knowledge that running IDLE Tasks consume nearly no CPU time. Of course the overall performance will be SLIGHTLY lower because of context switching and the time it takes for the idle process to finish its time slice (no it wont preempt after 1 op or something - will do a few usecs of processing till the OS notices something else has to be done)...
The real question, which hasnt been answered on that article is how much network bandwith does it consume? I'm running folding@home on a few machines here but never really had the time to check how much of our network bandwith its taking away...
Hopefully not that much
ps.: awake for 32 hours, this posting might not make sense at all
Before you email me, remember: "There is no god!"
Folding@home is a good client to use, it is so versitle. You can run it on nearly every OS out there and like the article states it is pretty much transparent. Plus it is easy to hide on a PC as a service. It seems to run on nearly no resources. It would be the smart choice for anyone. 'Nuff said.
[n8.r0n] http://petesweb.spymac.net/
...there's more than one reason it's called "Folding@HOME" and not "Folding@WORK". Hmm?
If you can't get permission the first time around, repeated pestering will not help your case.
Hokey statistics and ancient misconceptions are no match for a good thought in your head, kid!
It's the increased power consumption I'd be worried about in his (the bosses) position.
... which SETI@home didn't bother to inform people that their data distribution method wasn't exactly working. (Everyone was checking over the same exact patch of sky 24/7 for weeks). That's what initially turned me off to SETI@home, and I haven't been back since.
... it applies to Folding@home as well. Thousands and thousands of CPUs, running floored will eat up a considerable amount of power. Is it wasted? No, I don't think so, as long as the distributed computing applications are worthwhile and advance our knowledge.
While 1 CPU running at full throttle 24/7 isn't going to make that big of a jump in the power bill. 500 CPUs... 1000, etc... will create a huge increase in power consumption over a long enough time frame.
I fully support distributed projects like Folding@home, SETI, etc... and run them on my machines, both at home and at work, but the power consumption is a legitimate concern.
I believe someone did a (unprofessional) investigation of the SETI@home debacle when it first came out, and came to the conclusion that something on the order of 100 or 1000 barrels of oil per day were wasted on checking over the same data repeatedly
Regardless
Since most of these types of apps rely on "spare" CPU cycles, they basically keep the machine running at 100% cpu at all times. This causes the system to pull more power. Not only does this increase the electric bill, but it also keeps the temperature high all the time. This could have an impact on AC cooling costs as well, not to mention CPU life.
To me, that's the biggest deterrant from using it. I had been running the UnitedDevices client on my home computer. Since my computer ran all the time, I figured what the heck. But lately I've been trying to cut back on my power consumption. By leaving the UD client running 24/7, its like leaving an extra light bulb on, power wise.
A slip of the foot you may soon recover, but a slip of the tongue you may never get over. -Benjamin Franklin
OK, so I don't really believe the last one because it seems that most buildings have such variation in number of computers and people moving through etc. But that doesn't stop the other two for being equally accurate.
They have a message board (phpbb!) and not only is there a knowledgable community, but the admins usually comment.
http://forum.folding-community.org/
-- -- --
Help my mini cause: My journal
My company is moderately sized (~140 employees) and uses a large amount of bandwidth on a near constant basis. While our data does not need to flow in real time, any disruption in our network can cause quite an uproar directed at the IT department, of which I am a member.
While the data transfers involved with projects like SETI@Home and Folding@Home are small in comparison to our normal traffic, my superiors were concerned that if many connections were made to the central server simultaneously, there would be a noticable drop in performance.
I think this bandwidth issue, and not client performance, stands as the major roadblock to more corporate participantion.
-Shadow
...but it might help you get to the truth about why you're not being allowed to do it.
We all know that the vast majority of CPU cycles are wasted. If your boss is telling you that you can't do it because of the impact on the workstation, they're most likely lying to you. Most bosses either
Of course, addressing these issues with your boss is far from easy, but if proving to them that workstation performance is not the issue forces them to raise the real issue then at least you have a chance.
How can we afford to ever sleep
So sound again
--ebtg
i work for a canadian federal government department, and the main reason software like this is a huge no-no is mainly because it could have adverse effects on other in-house developed applications that are mission-critical to our clients needs. extensive testing is done only with "standard" applications that every user has.
we learned this the hard way when we thought installing w2k service pack 2 was a good idea when sp1 was the department standard. one of our in-house apps was crashing at random times and the suits upstairs were starting to ask questions. luckily this didnt come down on us as another problem was the cause of the crash's, and saved us a world of grief
Does anyone really think that the reason these things are being rejected by management is because of performance???
Like woodworking? Build your own picture frames.
Going off of the point that actually running these distributed programs are not free. I think these programs should allow companies to be given some sort of tax credit or break.
This would definitely speed up development, and provide an incentive to the companies with massive amounts of unused computers usually left on anyways during the evenings. At least at my work place this is the case.
I refuse to participate in a project that puts USofA in the center of the world map. Obviously, these are meant to degrade us who live outside this country.
Bring me back the good old Alaska-to-Siberia map.
I don't know if this is widely known yet but at work we have the google toolbar installed on our windows 2k workstations.
:-) There is also mention of being able to participate in other such distributed computing projects in the future.
The MIS guy at least approved their use.
Last week, I saw that the Google Toolbar had self updated, and one of the new features was the ability to opt in for participation in the Folding@Home project through the use of the Google Toolbar.
It appears that at the time this feature is limited to only a select clients. Nevertheless I sent a request to the MIS guy about it, and if I could enable it. He had no issues with it. (Aka run it if you want)
Perhaps if the MIS/IT person already lets you use the Google Toolbar on the Windows machine, then they would probably be more trusting of running Folding@Home through the Google Toolbar.
I haven't noticed any significant slow downs using regular mode, and in any case you can switch between regular and conservative modes. Conservative mode running when you're not using the computer.
Also although I dont have the link at the moment handy (at home on my Mac
.... ... }
int main (void) {
The report covers Windows, but I want to know about the effect on Linux. A couple of years ago I ran Seti@home on some of the Solaris boxes where I worked. Even though it was nice -19, it had a very noticable impact on system performance. My solution was a script that monitored the system load and killed the client whenever the server was busy doing real work.
In Windows, I think there are scheduling classes such that a low priority idle task will not receive any cycles if a normal priority task needs to run.
-Aaron
This post is encrypted twice with ROT-13. Documenting or attempting to crack this encryption is illegal.
A Duron 1.2GHz with 256 MB of RAM is a low end system? That's a pretty decent low-end system of *now*, but what about using a machine that's 2 years old or more? You know, those sub 1-GHz machines and 128MB of RAM (if you're lucky)? Man, that low-end system is far faster than what I use at work (and what most people use).
What about memory consumption? Having to hit the swapfile more often because its running would slow down a compile job, or heck, just the apparent responsiveness of the system. If opening a document takes 10 seconds longer because the system has to swap, I'd say that has a far more annoying impact than the miniscule extra CPU resources...
Exactly, even if the article is technically sound, factual, and clear. Management doesn't work on these principles. :-)
Seriously though, I cannot even run Linux on my desktop, because it is not "company approved software". I work for a big company, can you tell? When I even hint at anything Linux or Open Source, I get an immediate brick wall. No matter how logical or technically fantastic a solution, if it ain't "company approved", it ain't happening.
My beliefs do not require that you agree with them.
The problem with this is, there is no way that the wear on the machines is also "negligible." An average business workstation probably has something like 1% CPU usage average each day. When you bump that up to 100% (and drive and memory a related amount), it will shorten the computer's life.
This is why volunteer distributed computation has been primarily popular among academics, students, and low-wage tech workers; people who aren't financially responsible for the computers to which they have access.
Karma: Good (despite my invention of the Karma: sig)
True story, with details ignored / changed to protect the guilty:
A production system handling multi-million dollar transactions began to slow and crash for absolutely no reason we could fathom. As each degradation of the system was costing the company involved tens of thousands of dollars at a shot, the president and other higher-ups were growing quite irate about the difficulty and wanted it fixed NOW.
A few days of frustrating troubleshooting by a team of techs discovered the problem: Someone had installed SETI@Home on the production system and it was interfering with other operations. Having incurred hundreds of thousands of dollars of opportunity-cost losses in those few days, the 'professional' responsible for that stupidity was thrown out on their ass the instant they were identified as the culprit.
The moral of the story: If you're paid to be a professional, be one. Use business systems for _business_ and if you want to run fun stuff, do it on your own time and on your own dime.
I thought @home already folded!
You're using her as bait, Master!
If you're going to install something on Windows as a service, you're the system's administrator. You therefore effectively stop users who are not also administrators from playing with it, stopping it, etc. This would be directed at sysadmins who want to use the idle cycles of the desktop machines they administer, I believe.
Running something in StartUp is going to give you an application running under the logged-in user's context, that disappears when he logs out, and again can be fiddled with by said user.
We've had this problem with the GUI coders (VB and .NET monkeys) at our place. The stuff they produce will crash at the slightest provocation (such as being installed on the 'wrong' partition!). They then try to push the blame on to users for having 'non-standard' configurations (like more than one hard-drive). If someone can't write portable, maintainable, reliable, efficent code they should find a job that demands less technical skill - like flipping burgers.
By comparison the seti/folding@home clients are written to be portable (especially folding@home) and have been tested on many thousands of different computers, reliablity at the users end is more indicative of in-house issues.
It's a security risk, plain and simple. Running this on any company machine containing files that we care about, or that is behind the firewall, is too much of a risk to even consider.
Overly anal? No. All it takes is for someone to discover a buffer overrun in the application, create an exploit, and poison our DNS to get data from their site instead of folding@home's site. This is perfectly possible, and should it happen, could be devastating.
I don't care enough about folding@home to risk company security. The CPU cycles we would have spent crunching data for them are not an issue, especially if the cycles would have been wasted anyway. I would gladly spend those if there were no risk.
In any modern operating system, a "low priority" thread will happily take 100% of the CPU if nothing else is running. Low priority doesn't mean it hangs on 10% just in case something wants the CPU -- it means that if a high priority and a low priority process both want the CPU, the high priority process is going to get a larger slice.
nobody cares about your company. think about it. its important to you, but it makes a grain of salt of difference to anyone else.
Except every script kiddie out there who has a different, specific target in mind for a DoS attack. Most people hit by viruses and worms weren't even known to the attacker, let alone the intended target. If all the @home distributed computing projects are important enough to you, fine. If your security is more important, you'd be stupid to run the risk.
You don't get my point. If you have a P2P application or other process that is running and you actually want the application to take priority in all cases.
From excellent karma to terible karma with a single +5 funny post...
Extra heat from the CPU can add an extra ~5% to the cost of electricity required to run an air conditioner.
[I'm a mechanical engineer and have done some airconditioning design work]
Actually, computers can have a significant effect on air-conditioning services of buildings, particularly large computer labs which are in constant use (ie. computers are pumping out heat continually). I would not be at all surprised to find out that the running cost of airconditioning for a building would increase as a result of significantly more computers (or, as in this case, the computers pumping out significantly more heat).
"Because it's there." - George Mallory, when asked why he wanted to climb Mt Everest, March 18, 1923 (New York Times)
In today's dismal health arena, I wonder what the patent rights are on the results of such computing. They are using our CPU resources to find an answer and, even though the results might be published academically, some pharm company is going to take them and make a few over-priced drugs.
Perhaps someday I won't even be able to afford the drugs that are a result of my CPU cycles. That's not to discourage donating cycles, but it is something to think about.
Your boss doesn't know that it poses no risk to his property, and he is not obligated to listen to arguments to that effect. To him, taking the risk offers no return, so it makes no sense.
When someone might yell at me, it has to be OpenBSD.
Yes of course :)
But I prefer to download 2 gigs+ of extra pr0n than searching for Mulders sexual Partner....!
Before you email me, remember: "There is no god!"
Is the donation of time / resources to this considered tax deductable? That might sway a few PHBs.
A bit of a stupid question I guess, but does anyone know what the IP ramifications of this project are? I don't really want the spare cycles of my machine to help someone get a patent to lock the general public out of benefitting from this biz. I checked the page but I find no mention of IP, which hopefully means it's public material, but I'm not sure.
Point two is ridiculous considering how long the lifespan of cpu's are, and even more so if you take into account the fraction of that lifespan that they are actually in use, on average.
"Gold still represents the ultimate form of payment in the world." - Alan Greenspan, 1999
"similarity" in finding a cure for cancer. post a reply or email me to clarify, b/c I think I can answer your question.
a quick bio summary:
There are about 35,000 genes in the human genome, which means there are >35,000 different kinds of proteins in our bodies over our lifetime. Each of these proteins has a 3-dimensional structure that is nearly impossible to predict from genetic information alone. The 3-dimensional structure of a protein, along with its composition, determines the functionality of the protein. Determining the 3D structure of a protein and discovering the steps necessary for a peptide chain to wriggle up into a mature protein is called the "Protein Folding Problem"
In many kinds of Cancer, genetic mutations have occurred that cause either a problem with the way a protein folds up and thus changed its functionality; or a mutation has occurred such that the genetic instructions have changed, causing some proteins to be made more or less often than usual. Usually what happens is a whole lot of things get changed before cancer is diagnosed.
Those who can, do. Those who can't, simulate.
usually when idle, operating systems tell the cpu to sleep. in x86, this is called HLT. when it sleeps, it uses less electricity, it heats up less, etc...
(does not apply to windows 95/98, they dont HLT at idle. if you're using either of those systems, there are third party programs to HLT at idle time, though, to cool down the cpu.)
-- Matti Nikki
You're joking, right? You're comparing an OS upgrade like SP2 to running another application on your PC? Microsoft's half-assed "service-packs" aren't just software programs, and they do tend to fuck everything in the OS. Running other apps concurrently with whatever you have developed is a completely different scenario.
"I assumed blithely that there were no elves out there in the darkness"
What about the graphical version?
I prefer it over the text version. Its nice to look at and goes in my taskbar, but sometimes it will prevent 3d apps (games) from runing on my box. It seems to steal the focus from the games.
GF3 on dual p3 900's win2ksp3
my associative arrays can kick your hash - TCL
Folding@Home won't run on any of my Debian Unstable machines. It segfaults shortly after execution. Haven't had the time to track it down yet. Right now I am only running Folding on Debian Testing.
-- 4 8 15 16 23 42
Stealing? As so many Slashdotters would tell you, it's merely "Copyright infringing" on the electricity...
:)
oh wait
They certainly aren't. The system's "idle loop" these days often involves use of the HLT instruction, which powers down most of the CPU until it recieves an interrupt.
Making a fuss and planting trees on the weekend is great, you should also carry that presence of mind into your everyday actions.
Turn off your monitor when you leave for the day.
Don't not running hot water if you're not using it.
Thinking twice about having ten old computers on 24x7 in your house. It might be cool, but maybe DNS doesn't need it's own server (whoa, really?). Two one 24x7 is probably more than enough, unless you have some special need.
Sure there may be exceptions to these things -- that's fine. Just be conscious of your actions. Decide to leave the monitor on, don't just be lazy about it. (same thing goes for speeding, yelling at people, getting drunk, etc. They may or may not be bad, but decide to do them -- don't just fall into it.)
Maybe I'm some Berkeley nut-case, but usually when I leave work at night, I go by and turn off all of the monitors on my way out. People sometimes talk to me about it and I give them my shpeel, but when they leave the next day, it's more of the same. Is it too much for people to think about these things?
there is no thing
what else could you want?
Good point... didn't even think of it from that angle.
On my Windows 2000 box, I know my CPU is consuming more power when I run the client(s) because my CPU temp goes up when I run it, and down when I stop it. I know that windows has a lot more cruft than a Linux box, but overall it won't sit at max power while "idle".
A slip of the foot you may soon recover, but a slip of the tongue you may never get over. -Benjamin Franklin
It uses unused CPU time - it's supposed to occupy CPU cycles that you aren't using.
The fact that it only snags 50% CPU when everything else is turned off means that it must be single threaded - it's only using 1 of your Mac's two processors.
You can't use process viewer to determine how much it is stealing - the real test is to run another higher priority CPU intensive program and see how close Folding@home gets to zero.
Here's a CPU intensive program for your review (I assume you're running OS X).
I used to leave my computer on 24/7 before I bought a hardware gateway for my roommates to share net access. If I left my bedroom door open at night, it was fine, but if I had closed it, I'd wake up and notice my room is abnormally warm.
Since I've been leaving the computer off lately (using the nifty Hibernate feature), my room stays about the same temperature whether the door is open or closed..
And that's just one PC. I recently purchased a kill-a-watt device. I think I will give it a run and see the power consumption difference when I'm running a number-crunching background program vs a regular idle... I'll post results if I remember (feel free to nag me if I don't) =)
Please consider making an automatic monthly recurring donation to the EFF
I had a P4 machine that would slow down while doing number crunching. Turned out it was over heating when diddling numbers. The P4 chippy would slow itself down when it got hot. The solution was a bigger heatsink on the P4 and an extra case fan sucking the air out of the case, blowing it out the back. CPU temp dropped from 60C to 30C when running this app.
Well, of COURSE it uses 30 - 50% CPU time when the machine's idle! Heck, I'm running it right now and it's 'consuming' up to 90%...
that I wouldn't have used anyway. Folding@Home will ONLY use time that the processor would have otherwise not used at all. Gaah, stupidity.
TANSTAAFI: There Ain't No Such Thing As A Free iPod.
# Extra heat from the CPU can add an extra ~5% to the cost of electricity required to run an air conditioner.
How about the cost of processors dying because modern CPUs/cooling systems are not designed to run under full load full-time. We've had nasty problems with Athlons running these kinds of apps.
The #1 reason your boss, or your IT dept, has for turning you down has nothing to do with performance, and everything to do with stability (Talking about Windows here)
Simply, us IT folk generally don't want anyone running anything they don't NEED to run on a daily basis on their computer, period, because every additional thing adds complexity.
It's a stability and a security risk.
If you can't already explain process priority to your boss, you're fired.
However, if your computer is a laptop that you run on batteries for a significant fraction of the time, be careful - NiMH batteries really don't like to power CPU-burners, and as they age, they tend to fail in ugly ways. I used to have a one-hour train commute, and my laptop simply did *not* like running GIMPS. Also, even if I turned it off when I was running on batteries, it slowed down the recharge process significantly when I plugged back in again, and I don't think it liked that either.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Arrrghh...
Let me be more clear... I do not want any application to take priority over it! I want to be able to give max processing power to the collective, without any P2P app taking over many CPU cycles on my system.
From excellent karma to terible karma with a single +5 funny post...