I have Boinc (Seti & climateprediction) running on a machine that has now been delegated to running completely disconnected from the internet, on it's own internal network. I periodically connect my laptop to that network to manage and update it. Is there some way I can connect to the orphaned Boinc system to update work units, then reconnect to the internet to download to the project sites?
regards,
Dar7yl

The machine that downloads must process the WUs or they will be rejected.....

---

BOINC Wiki

The machine that downloads must process the WUs or they will be rejected.....

Does that mean that I would have to develop my own proxy server which acts like a Boinc server to the client, and a client to the real Boinc server?
Is there not someone who has already done this, perhaps to support clustering?

regards,
Dar7yl

---

The computer that downloads the workunits gets a host ID written on the server that is uploading the work to that computer. So the computer is known to the server.

If you would crunch a unit on this computer, then transfer the result to another computer for it to upload it, the computer uploading it would have the wrong host ID number the unit was sent to.

So what you can do is download work with one computer, move all Wus onto a CDRW or USB stick, upload them into the computer that will crunch them, then put all the results on the CDRW or USB stick again and let the computer that downloaded them upload the again.

Yet it might be easier to give the other computer a network connection of its own with internet access.

---

Does that mean that I would have to develop my own proxy server which acts like a Boinc server to the client, and a client to the real Boinc server?

What you're describing is pretty much the good old SETIQueue functionality.

Is there not someone who has already done this, perhaps to support clustering?

Every now and then a discussion around this topic emerges, but to date no utility has been forthcoming.
Once you consider all the implications it gets quite complicated and I expect that people just find it easier to install a gateway between BOINC and the Internet, be it a proxy or shared connection or something like that. Then they use the built in proxying and caching functionality of BOINC.

---

ralic's law of forums: Irrespective of any prior research done, you will find the solution to your question shortly after posting it to a public Internet forum, resulting in readers concluding that you have done no research on the matter whatsoever.

I too am interested in something like this.

I had a similar problem when I rebuilt my Rock Linux Cluster. Originally, I played around with the frontend, and eventually got it to where boinc was distributed out to my nodes with the project information and start automatically. When I had all of the nodes setup to see the internet this was no problem.

Now I am doing a little research work, and although it is not imperitive to have tight security, it just makes everyone feel better. These nodes sit idle 95% of the time. In this setup, I have nodes that can not directly see the internet, and the internet can not see the nodes. To the whole world, it is one computer when in fact it is a cluster of 5.

However, I never was able to get it to work in this new setup. I eventually was able to get boinc to think I had a massive lot of processors, and send my frontend a lot of work but I wasnt able to get the work farmed out to the nodes. Manually sending the work over to the nodes via a script resulted in all sorts of problems that left the frontend with more work then it could do in the time alloted. I gave up when the nodes were complaining about not getting any work, and the front end was completly bogged down.

I have been reading up on other boinc project sites, and I completly understand the fact that they do not like sending out a couple thousand requests to a computer, and having someone store up a BIG bunch of files that they dump on the main server all at once. I do think though, that it would not be hard for a someone to setup a cluster type version of boinc that downloads a certain amount of work units, farms them out to nodes, and returns them in a timely manner. My system (though flawed) does have promise in doing this and I have a few ideas that I will play around with later, but right now I just dont have the time to mess with it for more then a couple hours a week. So until then, my processors are sitting idle...

If anyone has an idea, let me know, and I will try it out when I get a chance. Who knows maybe a Rocks-BOINC Cluster is soon to come...

I too am interested in something like this.

How about a cluster of 184 hosts, sporting 370 CPU's ?
Read all about Merlin here http://pandora.aei.mpg.de/

Merlin has sucessfully run BOINC (Well it occupies top spot over at Einstein... :) Perhaps you can learn a trick or two to make your own cluster operate.

---

ralic's law of forums: Irrespective of any prior research done, you will find the solution to your question shortly after posting it to a public Internet forum, resulting in readers concluding that you have done no research on the matter whatsoever.