| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Support
Forum: BOINC Agent Support
Thread: Semi-offline computing |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 6
|
|
| Author |
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi everybody,
I work at IBM and I'm trying to get a project working with a cluster we have. I have access to an iDataPlex cluster that I have configured. Basically, there's a management node that has access to the internet and the cluster network, and the remaining 61 compute nodes are only on the cluster network with no access to the internet. Because the compute nodes do not have access to the internet, I don't believe there is any way for them to download work files. This is the mechanism that would be ideal to have working if possible: 1. Management node downloads work unit from internet, saves to network share 2. Cluster nodes process work unit from network share 3. Cluster nodes post completed work results to network share 4. Management node uploads completed work results to internet Is there any way this could work? The computing power that this cluster could provide, I believe, would be great for the WCG and I would hate to see it wasted. Thanks for your help! |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Hi,
----------------------------------------The man to mail is FAO knreed at support@worldcommunitygrid.org also at IBM. He or another WCG person will be able to head you further on in the right direction. cheers. --//-- edit: insert "direction" [Edit 1 times, last edit by Former Member at Aug 23, 2011 12:46:07 PM] |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges: 
|
Definitely send us an email at the support email Sekerob sent us.
Quick question here though. Is there any chance that you could set up a proxy on the management node? We can give you the only IP addresses that would be enabled. This would make it a very simple problem to address. You could even set up some file caching to reduce bandwidth if you wanted. |
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
Thanks for your support.
I have sent an email with further questions to the support address, and I'm currently in conversation with IT security. They were wondering how exactly the proxy should be set up in order to allow for the compute nodes access the WCG servers. In the meantime while I'm sorting this out, would you mind sending me the IP addresses of the servers that would need to be enabled in the proxy? Thanks again, Michael |
||
|
|
knreed
Former World Community Grid Tech Joined: Nov 8, 2004 Post Count: 4504 Status: Offline Project Badges:
|
Micheal,
I've sent this to you directly via email, but since others might benefit from this info as well here it is: When you talk with the security guys, make sure that they know the following:
As far as the proxy server goes, here are the ip addresses that you would need to allow access to: 198.20.8.246 198.20.8.241 170.224.194.69 170.224.160.205 170.225.97.195 The one exception to the IBM data center, is that we will upload a very large result file for The Clean Energy Project directly to the researchers at Harvard. The IP addresses for that are 140.247.231.147 140.247.231.157 If you choose to not allow those last two, then simply do not opt into The Clean Energy Project. |
||
|
|
martin64
Senior Cruncher Germany Joined: May 11, 2009 Post Count: 445 Status: Offline Project Badges:
|
Semi-offline crunching on BOINC is something I am doing. Here's how:
----------------------------------------I set up BOINC on the computer not connected to the Internet. Just download the client and run the script. This will extract the required files. Then I run boinc on the machine, set it up and stop again. This sets up the basic parameters. Next I copy all the boinc directory to the machine connected to the Internet. I run boinc on that machine, from that copied directory. I set up WCG as the project, and download enough WUs for a couple of days. I make sure that the host doesn't get tagged "reliable" in order not to get any WUs with a short deadline (by making the cache large enough). Then I stop boinc again. Move the directory back to the other computer, run boinc. And after a couple of days, start the process again. I think you could do that with a cluster as well, with an automated script as a workaround. The drawback is that all clients will have the same name, i.e. the name of the computer connected to the Internet. Regards, Martin  |
||
|
|
|