Attention Crunchers.
27 Mar 2005 20:03:24 UTC
Topic 188767
(moderation:
I was just sitting there reading logs, minding my own business, when I noticed that the CPDN project died last night.
I had an epiphany.
Why not start a BOINC project? One that will utilize spare processor cycles to accurately predict which BOINC project goes down next and when the outage will occur?
BoincDeathPool@Home
The Linux client will be written in compiled BASIC and utilize copious delay loops to further slow things down. This will ensure a lively message board.
The space-time continuum crunching should not be restricted to just BOINC project outages. It can also be used to predict exactly when the first whiney "We never had these problems with BoincDeathPoolClassic!" posts will appear.
Questions? Comments?
Attention Crunchers.
)
Nice idea!
Other things to predict: how long the outage will last, and what will cause the outage (hardware problems, software bugs, faulty breakers, ISP problems, cosmic rays, sunspots, gravity waves...)
The screensaver will show the probabilities of the projects going down, in nice colors (like red = "probably down at this point" and green = "likely to be up"), for a period of time equal to every project's WU deadline (so CPDN predictions will take most of the time). Of course these are the probabilities predicted by your set of parameters (workunit), another workunit will predict other probabilities.
The screensaver will compare the predicted outages to the real outages. (Care should be taken that this data-collection does not influence the outages of the other projects.)
Every BoincDeathPool@Home user who accurately predicts an outage (or aspects of it, like the cause) will be mentioned on an "honor roll" on the BoincDeathPool site. Some kind of metric to measure the accuracy of the predictions needs to be developed for this. The honor roll is in addition to the standard "cobblestone" stats and is an extra incentive to join this project (you will be famous on the entire Internet, not just a co-author of a hard-to-find scientific article).
The users that accurately predict an outage of BoincDeathPool itself will have a special place on the honor roll.
Oh, by the way, the Windows client should be written in Java to ensure it is not too fast, either. The fact it is in Java will be mentioned on the site, but not the fact you need to download some kind of JRE to run it. This should flood the forum with "how to install" questions and a few messages about interesting experiments to install it on non-Windows platforms. Furthermore, this garantees different bugs on Windows and on Linux.
;=)
Like your sense of humour
)
Like your sense of humour fellas.
ROFL, good one... Only
)
ROFL, good one... Only thing; what happens when the BOINC project death pool project goes down? Especially if someone couldn't report back the WU that predicted it's own downage until right after it comes back up? :D
Oh, and
> Oh, by the way, the Windows client should be written in Java to ensure it is
> not too fast, either. The fact it is in Java will be mentioned on the site,
> but not the fact you need to download some kind of JRE to run it. This should
> lood the forum with "how to install" questions and a few messages about
> interesting experiments to install it on non-Windows platforms. Furthermore,
> this garantees different bugs on Windows and on Linux.
should this be restricted to the SUN java virtual machine, or should Microsoft java be allowed? If the latter, convince M$ that it's a benchmark to their OS and losing it could be a major selling points (more credits and hence ego boost) for free Linux over expensive Windows liscenses, so they can find all the programming tricks to their own JVM to make it run faster...things like dynamically changing this or that instruction into making a Windows API call to cut down crunch time... Interoperability or maintaining cross platform compatibility not an issue. Risking ticking Sun off in court not an issue, as M$ can afford the best legal beagals, and we all know they managed to over-turn Judge Jackson's ruling, and got an appeals court to agree to them deciding their own sentencing in M$ vs. DOJ :D
Good thing I'm ading some
)
Good thing I'm ading some history thingies to BOS - then there'll be actual outage data to compare predictions to :D
BTW, the windows client should be made MS java compatible, but not sun - that will keep messages flowing on how to get ms java, how people hate ms and how unproffessional the project developers are :D
> BTW, the windows client
)
> BTW, the windows client should be made MS java compatible, but not sun - that
> will keep messages flowing on how to get ms java, how people hate ms and how
> unproffessional the project developers are :D
ROFL And oh yeah, just start taking benchmark scores (I remember having taken those SETI averages per platform type as a rough benchmark of how the CPUs stacked up, back in SETI classic days... True, each WU can very a bit in how long it took to complete, but when one gets a large number of WUs...and the numbers got to be very large, and takes an average time to completion with which to compare...)
Now, post these on Microsoft's usenet servers as indication of how Microsoft is getting creamed by Linux. Make a few comments about how Mr. Clippy (OK, that's Office, but peeps love to trash the sodding paper clip) and other bloat (generally bloatware) is slowing everything down. Use the benches as indication of how stream lined Linux is by comparison...and make sure to mention it in forums with an official Microsoft presense. Especially where PR or sales is the presence. Convince sales/PR this is a benchmark and how bad it makes them look.
Watch them acquire the project code (either publically if it's open source, or by reverse engineering/de-compiling..."though we don't officially talk of such things, per our lawyers suggestion") and inexplicably change their Java client with modifications meant to determine if this project is running, and if found, make M$ java VM modifications targeted explicitly to help this "benchmark", erm project get a higher score vs. competitor Linux :rofl
Following this, could come another Microsoft sponsored marketing study, targeted at the end user to "demonstrate" how much faster BOINC Death Project is on Windows (even in Java) then on that (whispered) "thorn in the side" Linux, and hence another reason the whole world should use Microsoft products for everything :D
> Watch them acquire the
)
> Watch them acquire the project code (either publically if it's open source, or
> by reverse engineering/de-compiling..."though we don't officially talk of such
> things, per our lawyers suggestion") and inexplicably change their Java client
> with modifications meant to determine if this project is running, and if
> found, make M$ java VM modifications targeted explicitly to help this
> "benchmark", erm project get a higher score vs. competitor Linux :rofl
I remember the days of the hobbiest magazines when they did a lot of benchmarking ... many compilers and video drivers were altered to recognize the benchmarking and would return higher numbers by such cheats ...
> I remember the days of the
)
> I remember the days of the hobbiest magazines when they did a lot of
> benchmarking ... many compilers and video drivers were altered to recognize
> the benchmarking and would return higher numbers by such cheats ...
Did they really do that??? my ... how childish :D
> > I remember the days of
)
> > I remember the days of the hobbiest magazines when they did a lot of
> > benchmarking ... many compilers and video drivers were altered to
> recognize
> > the benchmarking and would return higher numbers by such cheats ...
>
> Did they really do that??? my ... how childish :D
>
Alot of people still accuse some of the major chip and OS makers of doing that.
BOINC WIKI
BOINCing since 2002/12/8
why not give a try an use
)
why not give a try an use excel macros "compiled" for the *nix side?
There is always the risk of "datalost", it's just a "computer".
luck is coming and going away..
The other thing is: BACKUP!
With regulary backups of the cpdn (or other projects) Work, you will loose only until the last "save"
sad: the howto was written at so many places, there is no need to repeat the steps over and over..
> Alot of people still accuse
)
> Alot of people still accuse some of the major chip and OS makers of doing
> that.
Well, if they do ("they" being chip and OS makers) it's just plain disrespectful to their (potential) customers - potentially even shooting them selves in the foot (when people feel they've bought the cat in the sack they won't return)