improvement of BOINC


log in

Advanced search

Questions and Answers : Wish list : improvement of BOINC

Author Message
Profile bartsob5
Volunteer tester
Avatar
Send message
Joined: 16 Jun 04
Posts: 10
Credit: 6,715
RAC: 0
Poland
Message 222769 - Posted: 29 Dec 2005, 13:29:15 UTC

if there is anyone, who can do it, this is my proposition.

In my opinion BOINC should have an option, giving chance to hundreds of computers not being connected with internet. i'd like to propose that boinc should be able to read (on its own) data form for example FDD drive. Then there would be necessery only one computer connected to internet, and having BOINC-mother program, which would download workunits, newer versions of applications, and writing it to a FDD drive or flash memory. On computers which, are not connected to computer there should be also an option to export any data, if need (e.g. benchmarks, version of application) and of course results of WU being crunched.

Is it executable in general?
i'm looking forward for your opinions at this topic..
____________

Profile Pooh Bear 27
Volunteer tester
Avatar
Send message
Joined: 14 Jul 03
Posts: 3221
Credit: 2,218,412
RAC: 8,696
United States
Message 222786 - Posted: 29 Dec 2005, 14:32:18 UTC
Last modified: 29 Dec 2005, 15:25:19 UTC

This has been discussed several times. There are several issues here. Validaty, expirey dates, and general accounting.

Validating that a machine really did the work, and someone isn't cheating is harder in this instance. If someone wanted to make a farm of machines that just crunched more or less the same result, faking out the WU information, etc. would be a lot harder to catch (this happened in Classic, and is partially why this new system is in place).

Also stated on another thread about this subject is proving benchmarks. This can be cheated a lot easier when moving WUs machine to machine. The amount of data that would have to move back and forth would get humoungous with validaties in place. Then I still would think it would be easier to cheat.

Expirey dates could be missed a lot easier. If you are not diligent enough to move things back and forth every few days (which to me would get majorly boring after a couple of weeks), many results would expire.

Accounting for the WUs would be messed up. The information passed back and forth on benchmarks etc. are done only so often. How is the project supposed to know how to validate it speed wise, etc. when it can not keep track of the machine speed, etc. It would mean a major rewrite to try and keep this accounting in tact.

With the Internet as accessible as it is, it's not hard getting all machines to at least be able to send/receive the information from the projects. Using Proxies, personal networking, routers, etc. you can easily have all machines talking. You can block as much information as you want from the outside, if done correctly.

I really do not think this is a feasible solution anymore.


____________

Profile ksnash
Send message
Joined: 28 Nov 99
Posts: 402
Credit: 528,725
RAC: 0
United States
Message 224031 - Posted: 1 Jan 2006, 3:41:43 UTC - in response to Message 222786.

This has been discussed several times. There are several issues here. Validaty, expirey dates, and general accounting.

Validating that a machine really did the work, and someone isn't cheating is harder in this instance. If someone wanted to make a farm of machines that just crunched more or less the same result, faking out the WU information, etc. would be a lot harder to catch (this happened in Classic, and is partially why this new system is in place).

Also stated on another thread about this subject is proving benchmarks. This can be cheated a lot easier when moving WUs machine to machine. The amount of data that would have to move back and forth would get humoungous with validaties in place. Then I still would think it would be easier to cheat.

Expirey dates could be missed a lot easier. If you are not diligent enough to move things back and forth every few days (which to me would get majorly boring after a couple of weeks), many results would expire.

Accounting for the WUs would be messed up. The information passed back and forth on benchmarks etc. are done only so often. How is the project supposed to know how to validate it speed wise, etc. when it can not keep track of the machine speed, etc. It would mean a major rewrite to try and keep this accounting in tact.

With the Internet as accessible as it is, it's not hard getting all machines to at least be able to send/receive the information from the projects. Using Proxies, personal networking, routers, etc. you can easily have all machines talking. You can block as much information as you want from the outside, if done correctly.

I really do not think this is a feasible solution anymore.




How about this for a solution:
Computer is set up as cache host. It sets up client account caches. The client hosts cause encrypted accounting with cache host.

I put in an application to Berkeley. Hire me and I will show you how to do some of the stuff you keep saying you can't do.

I have had to turn off my computers network access because setiathome direct communications is giving me nothing but (500) http server errors and is causing a DOS state in my connection. Instead of having one time for one computer to connect and give results back, the many computers cause too much outgoing load so nothing goes out. I have a 1.5M DSL connection. 256kUp. I would like big brother to go away.
____________

John McLeod VII
Volunteer developer
Volunteer tester
Avatar
Send message
Joined: 15 Jul 99
Posts: 24391
Credit: 519,750
RAC: 26
United States
Message 226079 - Posted: 5 Jan 2006, 1:46:04 UTC - in response to Message 224031.

This has been discussed several times. There are several issues here. Validaty, expirey dates, and general accounting.

Validating that a machine really did the work, and someone isn't cheating is harder in this instance. If someone wanted to make a farm of machines that just crunched more or less the same result, faking out the WU information, etc. would be a lot harder to catch (this happened in Classic, and is partially why this new system is in place).

Also stated on another thread about this subject is proving benchmarks. This can be cheated a lot easier when moving WUs machine to machine. The amount of data that would have to move back and forth would get humoungous with validaties in place. Then I still would think it would be easier to cheat.

Expirey dates could be missed a lot easier. If you are not diligent enough to move things back and forth every few days (which to me would get majorly boring after a couple of weeks), many results would expire.

Accounting for the WUs would be messed up. The information passed back and forth on benchmarks etc. are done only so often. How is the project supposed to know how to validate it speed wise, etc. when it can not keep track of the machine speed, etc. It would mean a major rewrite to try and keep this accounting in tact.

With the Internet as accessible as it is, it's not hard getting all machines to at least be able to send/receive the information from the projects. Using Proxies, personal networking, routers, etc. you can easily have all machines talking. You can block as much information as you want from the outside, if done correctly.

I really do not think this is a feasible solution anymore.




How about this for a solution:
Computer is set up as cache host. It sets up client account caches. The client hosts cause encrypted accounting with cache host.

I put in an application to Berkeley. Hire me and I will show you how to do some of the stuff you keep saying you can't do.

I have had to turn off my computers network access because setiathome direct communications is giving me nothing but (500) http server errors and is causing a DOS state in my connection. Instead of having one time for one computer to connect and give results back, the many computers cause too much outgoing load so nothing goes out. I have a 1.5M DSL connection. 256kUp. I would like big brother to go away.

Or you can volunteer to write it. Berrkeley has nearly no budget, and much of the code was written by volunteers.
____________


BOINC WIKI

Questions and Answers : Wish list : improvement of BOINC

Copyright © 2014 University of California