LHC@home

I think it is time the entire BOINC server side community rethink the maximum amount of work a host can download and the duration before a WU times out and expires.

I run all my projects with a general preference of "connect to network about every .1 days"

Why?

1] I want to get the WU, crunch it, and return it post haste.

2] I don't want to hold up credit for anyone else because a WU will not process on my host for 10 more days.

3] I don't want to waste project's WU's if a "reset or detatch" becomes necessary for any reason.


LHC is just beginning to see the complaints about credits stacking up in the pending file.

Large host queues do nothing for the project. They are a capitulation to the volunteer hosts. I strongly ugre their elimination. Learn from SETI@Home's mistake: Large queues + short WU's = lost WU's, host complaints about credit not posting, validator backup, excessive db sizes.

LHC is in beta test, try a FIFO with the clients, see if it does not prevent many of the problems and complaints that SETI has had, and is continuing to have.

I believe you will discover less work is lost and your clients will be happier because credit will post sooner.

-P

I think S@H has trained people to be paranoid. They've started hording WUs to keep their machines busy during the frequent outages.

This is simply unnecessary now if you attach to multiple projects. Especially if you attach to CPDN. It's totally feasible now to run a small cache and always have plenty of work to do. :)

Well, everybody have not a direct and permanent connection to the net, or not reliable dialup connection without human intervention. So the computer must be able to work during relative long period (one week, perhaps two) without dialup. But the work is done, and the WU will be uploaded later.

These side effect are mandatory 'till boinc can't use a dialup connection.

This is the situation of one of my 3 computers who is in my parent's home.

Pete, you seem to forget that your situation isn't identic for all other , some people are in situations that force them to download for a long period. On my 2 other computer, I use the same modus operandi than you: download-crunch-upload very quickly...just BK I can do this!

Here my two cents of explication... AFAIK

> Well, everybody have not a direct and permanent connection to the net, or
> not reliable dialup connection without human intervention. So the computer
> must be able to work during relative long period (one week, perhaps two)
> without dialup. But the work is done, and the WU will be uploaded later.
>
> These side effect are mandatory 'till boinc can't use a dialup connection.
>
> This is the situation of one of my 3 computers who is in my parent's home.
>
> Pete, you seem to forget that your situation isn't identic for all other ,
> some people are in situations that force them to download for a long period.
> On my 2 other computer, I use the same modus operandi than you:
> download-crunch-upload very quickly...just BK I can do this!
>
> Here my two cents of explication... AFAIK
>
Your signature is too wide for my screen. Please edit it so that it wraps.

> This is simply unnecessary now if you attach to multiple projects. Especially
> if you attach to CPDN.

You have right, and not...
I run LHC on 50 PC, the Clients are all on the Server in a network share, i don't have the place (and won't the traffic) for the CPN Client. IF prediction stay stable i will add it, but in the moment i need to cache WU for 4 day's of LHC. i'm permanent online. I manually "flush" my buffers if they did'nt do.

Why i have all on the server is a longer story, but i think if i say "i cant write on local disk" it's enough info :-) [Automatic disk-restore on power on => lost all work if local]

Now i "wish" a personal proxy, with that i can distribute the Units, it's not perfect if two or four pc get every 10 monster-WU, and 40 or more run idle...
At the Moment most time in the 4 Day, it re-try all hour, it get some work in the four day, so most time they aren't idle.. but now my 50 PC buffer work for about 200 Days !! (50x 4 day) .. i'm unhappy if i catch workunits from you ;-), but i'm much more unhappy if the pc run idle.

> Well, everybody have not a direct and permanent connection to the net, or
> not reliable dialup connection without human intervention. So the computer
> must be able to work during relative long period (one week, perhaps two)
> without dialup. But the work is done, and the WU will be uploaded later.
>
> These side effect are mandatory 'till boinc can't use a dialup connection.
>
> This is the situation of one of my 3 computers who is in my parent's home.
>
> Pete, you seem to forget that your situation isn't identic for all other ,
> some people are in situations that force them to download for a long period.
> On my 2 other computer, I use the same modus operandi than you:
> download-crunch-upload very quickly...just BK I can do this!
>
> Here my two cents of explication... AFAIK
>
>
[/url]
</p>