When should Seti jetison the weak?

Message boards : Number crunching : When should Seti jetison the weak?
Message board moderation

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 5 · Next

AuthorMessage
PhonAcq

Send message
Joined: 14 Apr 01
Posts: 1656
Credit: 30,658,217
RAC: 1
United States
Message 890270 - Posted: 1 May 2009, 20:17:09 UTC

(Snazzy title selected to catch your attention and to incite your emotions.)

* There is some amount of fixed overhead at Berkeley for each host in seti. I won't speculate how much, but in aggregrate over the thousands of hosts it represents some amount (alot?) of work and resource consumption.

* The oldest, least productive computer hosts are contributing an increasingly inconsequential amount of work as time marches on. They are probably valued more for nostalgia and idealism, than for their bone-dry value to the project. Whatever, the ancient machines are consumming their overhead's worth of resources and contributing very little (in absolute terms) in return.

* There is a limited amount of resources to go around.

So I conclude that at some point, it would benefit Seti to stop giving work to the least productive computers. Have we passed that point with the 200K or so hosts?

If not, what number of P-II's and P-III's and so forth should the project support?

One answer to this is 'natural attrition', where the computers die, clients become apathetic, and the science application morphs into something the oldest machines can't support. That's fine, but I view that as a separate evolution mechanism.
ID: 890270 · Report as offensive
Profile KW2E
Avatar

Send message
Joined: 18 May 99
Posts: 346
Credit: 104,396,190
RAC: 34
United States
Message 890288 - Posted: 1 May 2009, 21:32:46 UTC - in response to Message 890270.  
Last modified: 1 May 2009, 21:32:57 UTC

That's kind-of like the Salvation Army telling people not to put any loose change into the bucket because it costs too much to count the change. Only bills please.

I understand what you are saying, and I agree to a certain extent. But I think the evolution of the app will kill off the old machines in due time.

Won't be too long before they double the work unit data size again. IMO

BTW, I'd love to see a 200k RAC machine.

Rob
ID: 890288 · Report as offensive
Profile S@NL - Eesger - www.knoop.nl
Avatar

Send message
Joined: 7 Oct 01
Posts: 385
Credit: 50,200,038
RAC: 0
Netherlands
Message 890294 - Posted: 1 May 2009, 21:52:02 UTC - in response to Message 890288.  

..BTW, I'd love to see a 200k RAC machine.

Rob

That way you'dd need to only maintain 1.65 machines you mean ? ;)
The SETI@Home Gauntlet 2012 april 16 - 30| info / chat | STATS
ID: 890294 · Report as offensive
Profile KW2E
Avatar

Send message
Joined: 18 May 99
Posts: 346
Credit: 104,396,190
RAC: 34
United States
Message 890298 - Posted: 1 May 2009, 22:05:00 UTC - in response to Message 890294.  

Wouldn't that be nice!
ID: 890298 · Report as offensive
Profile James Sotherden
Avatar

Send message
Joined: 16 May 99
Posts: 10436
Credit: 110,373,059
RAC: 54
United States
Message 890301 - Posted: 1 May 2009, 22:21:10 UTC

Me, I think attrition will get rid of the slow computers. I have and old P4 that takes 6.5 hours to crunch a WU. My mac will do 2 in 2 hours.Thats with stock apps. Sure i could do the OPT Apps. But i dont have the skill to install them yet. So when the IRS says yes to me, i will buy a quad i7 to replace the old P4.

[/quote]

Old James
ID: 890301 · Report as offensive
gomeyer
Volunteer tester

Send message
Joined: 21 May 99
Posts: 488
Credit: 50,370,425
RAC: 0
United States
Message 890332 - Posted: 1 May 2009, 23:26:43 UTC

Well, I don’t see a snowball’s chance of them “jettisoning the weak” and I’m not so sure it’s really a good idea in any event. Where would you draw the line? Who decides?

Only my opinion of course.

I would be satisfied if they just began refusing new work to 4.xx clients that give low or zero credits/wu.

I can't imagine how unhappy I would be if I had a workstation that takes 3 or 4 days to finish an AP WU and got zero credit for a perfectly good result just because my wingman is a cave dweller. Now that’s just plain unfair.
ID: 890332 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 890341 - Posted: 1 May 2009, 23:52:19 UTC - in response to Message 890270.  

(Snazzy title selected to catch your attention and to incite your emotions.)

* There is some amount of fixed overhead at Berkeley for each host in seti. I won't speculate how much, but in aggregrate over the thousands of hosts it represents some amount (alot?) of work and resource consumption.

* The oldest, least productive computer hosts are contributing an increasingly inconsequential amount of work as time marches on. They are probably valued more for nostalgia and idealism, than for their bone-dry value to the project. Whatever, the ancient machines are consumming their overhead's worth of resources and contributing very little (in absolute terms) in return.

* There is a limited amount of resources to go around.
...

My 200 MHz Pentium MMX system has an average turnaround time of 4.54 days. I've seen many much faster systems with longer turnaround than that, so they're causing more resource usage for workunit storage on the download server. Each system does have a host record in the database, but even hosts which are no longer active have that. What other kind of resources did you have in mind?

About 14 months ago the estimates and deadlines were adjusted such that a host needs a floating point benchmark of around 40 million ops/sec to get work. Perhaps Moore's Law suggests that should be about doubled every two years.

But my opinion is that slow hardware which is still producing valid work is not really more of a burden on the system than new hardware. If anything I'd suggest taking steps to eliminate hosts which don't produce at least 90% valid results, and that low a limit only because server-side problems can cause invalidations. Perhaps some method of dropping hosts and/or users who haven't contributed any results in the last x years would also be a sensible BOINC modification (x being selectable by each project).
                                                                Joe
ID: 890341 · Report as offensive
Profile Bill Walker
Avatar

Send message
Joined: 4 Sep 99
Posts: 3868
Credit: 2,697,267
RAC: 0
Canada
Message 890349 - Posted: 2 May 2009, 0:27:17 UTC - in response to Message 890270.  

snip snip

* The oldest, least productive computer hosts are contributing an increasingly inconsequential amount of work as time marches on. They are probably valued more for nostalgia and idealism, than for their bone-dry value to the project. Whatever, the ancient machines are consumming their overhead's worth of resources and contributing very little (in absolute terms) in return.

more snip snip


Your lack of concern for nostaliga tells me you are relatively young. Your time will come.

Your lack of concern for idealism causes me grave concern for the future of the species.

To make you a little happier, how about if SETI and BOINC charge the old machines twice what they charge you? Oh what the heck, three times.

But seriously, if Matt tells us they are lacking in total computing power right now, dropping the low productivity machines just means the big crunchers still don't have enough CPU for what SETI wants. Maybe you should offer to pay us little guys to upgrade our CPUs?

ID: 890349 · Report as offensive
Profile Blurf
Volunteer tester

Send message
Joined: 2 Sep 06
Posts: 8962
Credit: 12,678,685
RAC: 0
United States
Message 890397 - Posted: 2 May 2009, 1:56:16 UTC - in response to Message 890270.  

* The oldest, least productive computer hosts are contributing an increasingly inconsequential amount of work as time marches on. They are probably valued more for nostalgia and idealism, than for their bone-dry value to the project. Whatever, the ancient machines are consumming their overhead's worth of resources and contributing very little (in absolute terms) in return.


I find this to kind of be insulting towards people with smaller farms or older systems. As long as any computer is contributing anything I find it extremely difficult to call it "bone-dry value" that they offer.


ID: 890397 · Report as offensive
Profile Misfit
Volunteer tester
Avatar

Send message
Joined: 21 Jun 01
Posts: 21804
Credit: 2,815,091
RAC: 0
United States
Message 890403 - Posted: 2 May 2009, 2:31:21 UTC - in response to Message 890270.  

(Snazzy title selected to catch your attention and to incite your emotions.)

I thought you had misspelled week. SETI on vacation.
me@rescam.org
ID: 890403 · Report as offensive
Profile zoom3+1=4
Volunteer tester
Avatar

Send message
Joined: 30 Nov 03
Posts: 65640
Credit: 55,293,173
RAC: 49
United States
Message 890405 - Posted: 2 May 2009, 2:37:27 UTC - in response to Message 890397.  

* The oldest, least productive computer hosts are contributing an increasingly inconsequential amount of work as time marches on. They are probably valued more for nostalgia and idealism, than for their bone-dry value to the project. Whatever, the ancient machines are consumming their overhead's worth of resources and contributing very little (in absolute terms) in return.


I find this to kind of be insulting towards people with smaller farms or older systems. As long as any computer is contributing anything I find it extremely difficult to call it "bone-dry value" that they offer.

Blurf You read My mind, The project needs any capable computer It can get, Besides they do the work, Just really slow.
The T1 Trust, PRR T1 Class 4-4-4-4 #5550, 1 of America's First HST's
ID: 890405 · Report as offensive
HFB1217
Avatar

Send message
Joined: 25 Dec 05
Posts: 102
Credit: 9,424,572
RAC: 0
United States
Message 890430 - Posted: 2 May 2009, 4:29:19 UTC

Hell that is a poor way to treat members who can not afford to upgrade or do not need to upgrade because what they have meets their needs.

It is also a bit elitist after all this is a project for all who want to participate in the project. Who knows one of them just might find ET!!
Come and Visit Us at
BBR TeamStarFire


****My 9th year of Seti****A Founding Member of the Original Seti Team Starfire at Broadband Reports.com ****
ID: 890430 · Report as offensive
PhonAcq

Send message
Joined: 14 Apr 01
Posts: 1656
Credit: 30,658,217
RAC: 1
United States
Message 890501 - Posted: 2 May 2009, 11:11:21 UTC

I find it curious that no one thus far has based their response on objectively analyzing the argument I tried to outline. Let me ask again a different way.

If one-half of the 200K active participants ran P-II class machines and produced a result every week, I wondered if that was helpful.

There is overhead involved with keeping up the various databases, with keeping the long residence time wu's active on the servers that limit available storage for additional wu's, with backing up and resurecting this information during server failures, and so forth. Time and money spent doing this for the weakest hosts may be better spent on other tasks.

But if 100K P-II's were tolerable, then would 300K or 1000K be the breaking point? Obviously that depends on available resources. But I hope we can all agree there is an upper bound.

So my challenge to readers is to estimate (objectively) what this limit is.

I don't have access to the data needed to analyze the problem. For example, just how many 'slow' cpu's are out there and what percentage of the daily production do they represent? Or, what could be saved if the bottom X% of hosts were jettisoned? Surely, this type of data is buried somewhere. I would hope someone reading this thread would be able to post some of it, or there estimates of it.

And yes, I have a dual P-II 333MHz that keeps chugging and heating my house. It won't fail (as I had hoped a year or so ago so that I could turn it off for good), so I'm looking down this line of reasoning for another objective reason to turn if off.
ID: 890501 · Report as offensive
-ShEm-
Volunteer tester

Send message
Joined: 25 Feb 00
Posts: 139
Credit: 4,129,448
RAC: 0
Message 890515 - Posted: 2 May 2009, 12:31:57 UTC - in response to Message 890501.  

...I don't have access to the data needed to analyze the problem. For example, just how many 'slow' cpu's are out there and what percentage of the daily production do they represent? Or, what could be saved if the bottom X% of hosts were jettisoned? Surely, this type of data is buried somewhere. I would hope someone reading this thread would be able to post some of it, or there estimates of it...

Maybe boincstats's S@h processor-stats (sorted by average credit) can help?
ID: 890515 · Report as offensive
Profile Bill Walker
Avatar

Send message
Joined: 4 Sep 99
Posts: 3868
Credit: 2,697,267
RAC: 0
Canada
Message 890522 - Posted: 2 May 2009, 12:45:05 UTC - in response to Message 890515.  

...I don't have access to the data needed to analyze the problem. For example, just how many 'slow' cpu's are out there and what percentage of the daily production do they represent? Or, what could be saved if the bottom X% of hosts were jettisoned? Surely, this type of data is buried somewhere. I would hope someone reading this thread would be able to post some of it, or there estimates of it...

Maybe boincstats's S@h processor-stats (sorted by average credit) can help?


The fact remains that Matt tells us the choke point right now is CPU power, not anything on the server side. This is data accessible to anybody who can read and reason.

Jettisoning ANY CPU power will not change this. That is a rational, logical argument,that can be added to the previous posts about the original question's lack of humanity.

Sorry, I know it is just a troll, but us old guys get worked up easily.


ID: 890522 · Report as offensive
Profile elbea64

Send message
Joined: 16 Aug 99
Posts: 114
Credit: 6,352,198
RAC: 0
Germany
Message 890526 - Posted: 2 May 2009, 12:54:05 UTC

Objectively there is no limit because for every limit one can think of, there's a workaround. so nothing curious about non-objective answers

subjectively - means from a special point of view - we need the point of view you want to discuss

  • technic
  • money
  • users
  • staff
  • politics
  • project
  • and so on ...



In the end this discussion doesn't make much sense because of the lack of facts.

the only limit that will come was mentioned already, is the doubling of the work per WU, which will drop a lot of hosts that can't do a WU within the deadline anymore. When this happens the "old" hosts won't be purged from databases, so the overhead you're talking about won't be the reason for doing this. the most likely reason for doubling the work per WU will be the fastest hosts as faster hosts will hammer servers more often the faster they get and the overhead for doing all the administrative things for every WU will be less for greater WUs.

ID: 890526 · Report as offensive
Profile Bob Mahoney Design
Avatar

Send message
Joined: 4 Apr 04
Posts: 178
Credit: 9,205,632
RAC: 0
United States
Message 890535 - Posted: 2 May 2009, 13:08:26 UTC - in response to Message 890501.  

And yes, I have a dual P-II 333MHz that keeps chugging and heating my house. It won't fail (as I had hoped a year or so ago so that I could turn it off for good), so I'm looking down this line of reasoning for another objective reason to turn if off.

The "incite" portion of your plan worked!

This sounds like the tough decision I had to make a few years ago when I finally decided to limit my online system's support of 110baud and 300baud modems. People still used them, but maintaining compatibility with them was stifling to the overall system progress.

In the big SETI@home picture, I think that is happening here, maybe more than a little. Like, the micro-size of the MB WU vs. the mega-size of the AP WU.... well, one is too stuck in the past, and the big one needs a fast, modern computer. This makes the science side of SETI a bit schizy in appearance. It also makes the project less competitive with all the other projects competing for volunteer distributed computer time. That is, users of old computers hate AP, and users of hot-shot new stuff get tangled in a mess of micro-sized MB workunits that clog up the servers, so nobody wins.

I'm sorry to harp on saving energy, but doesn't anyone worry about the extremely ineffecient use of electricity with the very old systems? I've wondered if progressive thinking at Berkeley is in denial on this topic?

Hey, don't get too angry at my opinions..... I still worry about how much of an energy usage pulse the U.S. experiences when a national talk-radio host pounds on his desk during a broadcast - each of those bass pulses cause a lot of speaker movement, and multiply that by 5,000,000 listeners with two speakers each..... I just solved the energy crisis..... cut off the weak computers on SETI, and put pillows under the arms of Rush Limbaugh and Howard Stern!

Energy Czar, you say? Aw shucks.

Bob
Opinion stated as fact? Who, me?
ID: 890535 · Report as offensive
Andy Williams
Volunteer tester
Avatar

Send message
Joined: 11 May 01
Posts: 187
Credit: 112,464,820
RAC: 0
United States
Message 890542 - Posted: 2 May 2009, 13:45:06 UTC - in response to Message 890535.  
Last modified: 2 May 2009, 13:46:05 UTC

doesn't anyone worry about the extremely ineffecient use of electricity with the very old systems?


That is the best argument for retiring old systems. FLOPS per watt. You can do two things with a newer system. Generate a lot more results for the same amount of electricity ($) or generate the same number of results with a lot less electricity ($). In the latter case, the new system will eventually pay for itself. Personally, I have retired all single core CPUs.
--
Classic 82353 WU / 400979 h
ID: 890542 · Report as offensive
Profile Bill Walker
Avatar

Send message
Joined: 4 Sep 99
Posts: 3868
Credit: 2,697,267
RAC: 0
Canada
Message 890545 - Posted: 2 May 2009, 13:57:07 UTC - in response to Message 890542.  

doesn't anyone worry about the extremely ineffecient use of electricity with the very old systems?


That is the best argument for retiring old systems. FLOPS per watt. You can do two things with a newer system. Generate a lot more results for the same amount of electricity ($) or generate the same number of results with a lot less electricity ($). In the latter case, the new system will eventually pay for itself. Personally, I have retired all single core CPUs.


A lot of this makes sense, but surely it is the user who decides this, not SETI admins.

If, for example, an energy concerned user spends all their capitol making the rest of the household very energy efficient (new light bulbs, insulation, appliances, solar panels, etc.), and then can't afford to upgrade the old PC, his total energy use per person could be much less than somebody running a shiny new quad processor in a big house, with the windows open while the a/c runs, and with all the incandescent lights left on. The new PC could be replaced later, when the cash flow permits. Meanwhile, they continue to search for ET with a clear conscience.


ID: 890545 · Report as offensive
Profile elbea64

Send message
Joined: 16 Aug 99
Posts: 114
Credit: 6,352,198
RAC: 0
Germany
Message 890547 - Posted: 2 May 2009, 14:16:04 UTC

doesn't anyone worry about the extremely ineffecient use of electricity with the very old systems?

BOINC is intended to use not otherwise used cpu-cycles, which means that the old systems are running regardless if BOINC is running or not, so it's not BOINC that is inefficient in that case, it's something else and therefore no reason not to run BOINC on such systems.

So the question could be if it is inefficient to to buy/run systems only for BOINC which is the same argument for old and new systems, i mean even more for new systems

*just another point of view*
ID: 890547 · Report as offensive
1 · 2 · 3 · 4 . . . 5 · Next

Message boards : Number crunching : When should Seti jetison the weak?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.