hey thanks thats just what i wanted. people that would talk strate to me. one question why cant u rewrite the part of the seti software to either break the work unit up into muopital parts so you can use mupital instances or rearange the calculations so that you are maxing out the processors.

if any thing just make sure your using all the math resisters on the processor all
the time...

O you did't hury my feelings i just wanted a strate ansure. so thank you...

---

Hey the guy has a idea and its good but not achievable as things are. This was argued in a thread on on Von Neuman and Harvard to some extent. Parallel is right but the ahrdware and compilers that would take advantage of that does not exist in a production sense AFAIK. It we take out the synchronised element of parallel operations and view the processing as asynch streams then we achieve it with Boinc/seti....multiple processors (our PCs) take a small part of a big mathematical problem and sending back results for ultimate compilation. We are parallel now but in an asynchronous manner. Going further than this is not possible but, as agrued by me amongst others in the other thread, is the future for sure......but perhaps not in the next 10-15 years given how Intel and AMD are wedded to their current archtectures.

Of course its always open to him to get the seti code and see if the processing can be split down so different CPUs on different PCs take parts of it and do process it more in parallel. Not true parallelism but a new take on how the math can be done.

---

Im sorry if I was one that upset you. any contribution to SET@home is always welcome here.
I once tried my hand at setting up a Beowolf cluster. I got better results with boinc installed normally on each PC. It was only 7 PCs. If you run 100+ you may have better result.I have not been able to find published results from anyone trying it on that many PCs. I have followed numerus threads over the past year where others have tried parallel computeing in various ways and LTSP but all seem to come back to the same conclustion. you may come up with the answer that all of us/them needed. but you will only know if you try. So I'm behind you on this. There have been post of other's presently running parralell systems but in that same time none of the users of those systems have posted how they are doing it or any rateing of those systems. one I remember stated that he maintained a super computer overseas(europe somewhere) when they didn't have other work he ran seti on it.
I'm sure he didn't want to set it up as individual boxes in the off time then set it up for parallel computer just before doing work for a paying client so it ran as a supercomputer.
presently Boinc only runs 1 instance on 1 CPU at a time. if you have Hyperthreading CPU then you have to run multiple instances to max out the chip. same with Zeon. Maybey someone has already fixed this but not shareing.
I would suggest some searching. maybey team 2cpu could be of more help.

thank you for telling me that i am not a sofware enginerr i an a computer builder.
i know harw ware not software.

now sience i've got your attion i think i will try somthing else.

anser me this if we could hipethitey get a super computer for the seti project and
someone could rewrite the bonic software to sun on such a system. would you be interested.

you sound like you know alot more about the software than i do. but the system i was looking at to help seti at home would do around 460 gflops and was very good.

you say we have a tone of menbers do we? well we should start a fund raizer. depending on how meny people donate. we just might be able to pull it off.

if any thing can we get a better server system one that atleast is up more than 1 hour a day.....

---

<blockquote>you know you are all prety cocky. does a new idea scare you so much or theriten your stats so much that you have to bash it. her it is proven;
1. bonic can run on mutipul processores becouse it can run on a 2 cpu system and use boath processores.

I almost hate to get in this but since I have been encouraging you I feel I MUST help set the record straight! Boinc DOES run on multiple processor machines BUT it runs ONE process on EACH processor, NOT the same process on multiple processors!
Like adding 2 + 2, if I ask you and at the same exact time ask lots of other people the same question, you each go thru the EXACT SAME PROCESS to get the answer. You do not share the data to come up with the answer. Okay take the idea and expand it a bit....(2 + 2) + 5 / (3 * 43). Each person will come up with, hopefully, the right answer, but again no one shared the processing with anyone else. This is how Boinc works now.

You are talking about running the SAME process on multiple processors thereby speeding up the output. You are talking about taking that last math problem and giving you the (3 * 43) part, someone else doing the (2 + 2) part and someone else still doing the rest. Then each of you giving your individual results to another person to do the final calculations and come up with the right answer.
The Boinc program as currently written will NOT do that!

---

you know you are all prety cocky. does a new idea scare you so much or theriten your stats so much that you have to bash it. her it is proven;
1. bonic can run on mutipul processores becouse it can run on a 2 cpu system and use boath processores.

2. tlets do some math shale we. it takes a computer on average to do one work unit 4 houres to do a work unit. there are 24 houres in a day 7 days in a week.
so 24 * 7 = 168 houres. in a week. now to find how meny work units you can do on averige in this time we do 168 / 4 = 42. per computer now lets finish the equasion
to find how meny ten computers do in this time we do 42 * 10 = 420. pre week. now not including the constantly down servers for seti at home it takes 1 minit per work unit 30 sec to get it and 30 sce tosend the resoults. so the wasted time can be found. so there is a wasted 420 minits a week with just 10 computers wich equates to 420 / 60 = 7 houres which is aprox 2 work units. the cluster would not have this problem becouse it could get bigger work units and do them much mor quickly. also the speed loos for a os on every system that is curently running bonic is enormus. it equates to 10% to 30% of the processores actuaull work time avable. that is just for the os a clustor would be wasting around 3% to 10% which in a 50 computere clustor is a very large increase in speed.

o and i know all of you are supoesd to be computers neerds but not one of you has given a soultion or guiven a reson based in fact for why this is stupis and will not work. so lets see a little mor intelegentens.

>>Seems like it is time to start writing the code then! IF you can do it AND you have the resources, go ahead and PROVE that your way WILL work AND be more efficient. It just may become an add-on program for Boinc.

i am asking for people in the comunity for there help to advance our project. if that is such a bad thing to you maby you should quit. becouse i want to see seti at home flurish. also this could be applied to meny other projects.

amd another thing it is being done all around the world. it is just called diferint things. it is usen in meny lab's and in data crunching compineys.

---

<blockquote>>>But can the Boinc software be subdivided into multiple processors like that? I don't think that it can at present, meaning that you would have to write your own software and then try and interface it back together. May not be worth the effort, except for us "farmers". And even then there are MANY more single machine people than us "farmers".

yes the bionic software is capible of using muptial processores but how meny i don't know.

but the people that have been saying that it is not faster or just a little that the way you all do it now is incorect. just look at the latencies that take place when you run your software. in one week with just 10 computers running at an avwerage rate would wast about 1 to 3 houres. this of corse is not including when the server is down which is starting to look like it is happing dayley.

but also the reson i want to do this is not only becouse i want to help the seti project but becouse i have about 50 to 100 computers i want to put to work for this project and i did't want to have to futzs with installing and configuring that meny os's. also if i can make a working clustor i can make it take a lot less space it does now.

also a nother good thing about this system is it would be continuley be getting faster becouse i am continuley accuring new computers. and i know if i dont put this in hear somewone will reem me out for it so yes i know that there is a maxnum speed that this system will be able to go becouse of network speed limatations; so when that happends i will make another clustor, and continu on.</blockquote>
Seems like it is time to start writing the code then! IF you can do it AND you have the resources, go ahead and PROVE that your way WILL work AND be more efficient. It just may become an add-on program for Boinc.

---

Crash programs fail because they are based on theory that, with nine women pregnant, you can get a baby in a month.

Wernher von Braun
.

---

BOINC is aimed at single processors so, the only feasable approach is to just install BOINC on each computer attach S@H (and other projects) and let it do its thing.

---

BOINC WIKI

>>But can the Boinc software be subdivided into multiple processors like that? I don't think that it can at present, meaning that you would have to write your own software and then try and interface it back together. May not be worth the effort, except for us "farmers". And even then there are MANY more single machine people than us "farmers".

yes the bionic software is capible of using muptial processores but how meny i don't know.

but the people that have been saying that it is not faster or just a little that the way you all do it now is incorect. just look at the latencies that take place when you run your software. in one week with just 10 computers running at an avwerage rate would wast about 1 to 3 houres. this of corse is not including when the server is down which is starting to look like it is happing dayley.

but also the reson i want to do this is not only becouse i want to help the seti project but becouse i have about 50 to 100 computers i want to put to work for this project and i did't want to have to futzs with installing and configuring that meny os's. also if i can make a working clustor i can make it take a lot less space it does now.

also a nother good thing about this system is it would be continuley be getting faster becouse i am continuley accuring new computers. and i know if i dont put this in hear somewone will reem me out for it so yes i know that there is a maxnum speed that this system will be able to go becouse of network speed limatations; so when that happends i will make another clustor, and continu on.

---

> it has the problem is just finding people with the money and will to do it.
> you can do it with suse and a free app that will make the muptal mechines work
> as one with moutpal processors.
> this is done all the time insted of using traditinal super computers...
>
But can the Boinc software be subdivided into multiple processors like that? I don't think that it can at present, meaning that you would have to write your own software and then try and interface it back together. May not be worth the effort, except for us "farmers". And even then there are MANY more single machine people than us "farmers".

---

As I understand it, some versions of linux and/or BSD already support parallel computing by sharing processor usage through network connections. The advantage is that only one of the computers would need a full hard drive and supporting bits like that. The others that are simply offering processor power need no more than one or two floppy drives with a bare-bones OS on them to operate in the grid. There are limitations to this, however, because at some point the processing required to keep up the parallel computing outweighs the leftover capacity to do work. Theoretically, you could arange such a system and run BOINC on just one computer. However, for the amount of hardware that you would need to construct this, you're just as well off running BOINC independantly on all the machines.

---

>yes i under stand that but that is not as efficent or nearley as fast as a true >parallel computer soultion would be. if this was done right it would take minutes >to complete each work unit. thus providing a much faster and more reliableway to >to do the work plus it could be used to help other bonic project

Ok let me see here... A true parrallel system resides at a location with all of the PCs at the same location(Geological location). I can't argue that it is more efficient . Lots of money in funding BUT a GRID SUPERCOMPUTER such as us is a little different. The transfer of data between us may be slower but it flows and it is right. even though it takes longer. We have more people. And we grow every day.
take into account the number of PCs that the supercomputer has and the numbers that we have.

it has the problem is just finding people with the money and will to do it.
you can do it with suse and a free app that will make the muptal mechines work as one with moutpal processors.
this is done all the time insted of using traditinal super computers...

---

> yes i under stand that but that is not as efficent or nearley as fast as a
> true parallel computer soultion would be. if this was done right it would take
> minutes to complete each work unit. thus providing a much faster and more
> reliableway to to do the work plus it could be used to help other bonic
> projects...
>

If your Idea could work I am sure that would already be done........

Seems logical.........

---

yes i under stand that but that is not as efficent or nearley as fast as a true parallel computer soultion would be. if this was done right it would take minutes to complete each work unit. thus providing a much faster and more reliableway to to do the work plus it could be used to help other bonic projects...

---

The seti app is already about as parallel as it can get. Just hooking them up to the internet and running 15 copies of the client is the best way to do it.

---

BOINC WIKI

BOINCing since 2002/12/8

Hey Is any one famillear with Parallel computing. i would like to take the 15 or so computers i have laying around and set up somthing that can help seti at home and get more work units done faster. any help would be apericated.

---