| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: Supporting more than one project - Pros and Cons |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 14
|
|
| Author |
|
|
Dirk Gently
Senior Cruncher England Joined: Mar 1, 2005 Post Count: 153 Status: Offline Project Badges: 
|
I have always had LIM set to on. I have plenty of memory, amd swapspace. However if the other option now works, without any loss of processing, I think I will change it. But I am still not clear exactly how it works. Does a project switch force a checkpoint first? Would it also be better to increase the project switch interval to more than the default 1 hour? What time would be best?
---------------------------------------- |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
Why would it be best to Leave in Memory turned off?  If you're not severely memory-limited, it's better to keep "Leave in memory" turned on. For one thing, it seems if you're not crunching while active, you'll otherwise lose everything since last checkpoint. Also, even some applications like example SETI@home and earlier Einstein@home (not sure with current) can basically checkpoint once per second, some re-initializing is done each time starts from checkpoint, meaning you'll lose maybe 30-60 seconds each re-start... As for multiple projects, even if you're running CPDN and should in theory have work for some months into the future, you can still hit a bad wu that terminates at the same time server has an unexpected outage... Meaning, even if your 3.76-days cache seems just the right thing to handle any outages, you can still hit a string of "bad" wu's. So, having one or more "backup"-projects with a low resource-share is an advantage. That projects as backup is up to you to decide.   "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." [Edit 1 times, last edit by Ingleside at Jun 24, 2007 11:12:37 PM] |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
No, checkpoints can't be forced (unfortunately). But the redesigned scheduler will avoid switching until a task has checkpointed, and will use checkpoints as an opportunity to reschedule the CPU.
I'm not familiar with all the internal details (it's very complicated, and documentation is.... sketchy) - but that's pretty much the overview of how it behaves. I only run WCG, so I don't have any intuitive idea of what the best project switch interval is. Still, I recommend a larger value than the 1 hour default. Before the scheduler rewrite, I advised 6-12 hours. Pick what suits you best, and see how it goes. It probably is heavily dependant on how many projects you run. Of course, if you only run WCG, then the multi-project settings are completely irrelevant (although they may still be used in some scheduler calculations, but not any important ones). Am I the only one that finds it confusing having the CPU scheduler on the client, and the work scheduler on the server? Scheduler this, scheduler that.... |
||
|
|
keithhenry
Ace Cruncher Senile old farts of the world ....uh.....uh..... nevermind Joined: Nov 18, 2004 Post Count: 18667 Status: Offline Project Badges: 
|
Here's a bit of a different take on the original question. Whether you crunch for WCG only or multiple projects is really the same sort of decision as deciding how many and which non-profit groups you want to donate money to. It's all a matter of your interests, priorities and emotions. Once you decide on how many and which projects, then sorting out the best implementation of that decision is the next step. It's a lot simpler if you choose one project of course but crunching multiple projects on even a single machine is doable. It's the crunching equivalent of someone with little money to spare donating to a number of non-profits instead of just one. It's still a personal decision.
---------------------------------------- |
||
|
|
|