Project News - LHC@Home

[Cruncher Pete]

If you only run it on one machine and one unit at a time there's no benefit to running a network proxy just for LHC.  If you run multiple units and multiple machines, even if they're all VBox units, then the saving in network traffic, and the amount of time saved from not having to download common files over and over again, is significant.

Guys, I have to admit to something.  Currently I am running 8 machines and a number of mini machines.  I am an old bugger not just by age but from computer experience. I am just a push it and see person, have no idea what the best setup for anything is.  It is time to learn.  For example. please explain what a network proxy is as far as crunching is concerned. I might have been running my machines incorrectly, not set up at all for all I have been doing is install the beast, flick the on switch and hope it will do its thing.

Dingo. you might move this thread into a more suitable one than LHC@Home.

[Dark Angel]

If you only run it on one machine and one unit at a time there's no benefit to running a network proxy just for LHC.  If you run multiple units and multiple machines, even if they're all VBox units, then the saving in network traffic, and the amount of time saved from not having to download common files over and over again, is significant.

Guys, I have to admit to something.  Currently I am running 8 machines and a number of mini machines.  I am an old bugger not just by age but from computer experience. I am just a push it and see person, have no idea what the best setup for anything is.  It is time to learn.  For example. please explain what a network proxy is as far as crunching is concerned. I might have been running my machines incorrectly, not set up at all for all I have been doing is install the beast, flick the on switch and hope it will do its thing.

Dingo. you might move this thread into a more suitable one than LHC@Home.

I'd say leave it here in LHC as it's very LHC specific.

The proxy, specifically a caching proxy ('squid' is just the name of the proxy software), stores files you download locally (only works for non-encrypted traffic) so when your machines call for that same file again the proxy provides it instead of having to download it from the original site all over again.  LHC work downloads a lot of data from their servers for each work unit, not just the work unit itself.  Since your local network is typically much faster than your internet connection this means not only a reduction in the amount of data you're downloading each time a LHC work units starts but it downloads from the proxy to the Boinc client a lot faster so the unit doesn't waste as much time sitting around waiting.  The more machines and the more work units you have running at once the bigger the difference this makes to your network usage and overall efficiency.
Some other projects like WCG do NOT like being run through a caching proxy, so the system is typically set up such that it passes WCG traffic straight through instead of caching it.  HTTPS traffic isn't cached as it's encrypted.

There's information about setting it up here: https://lhcathome.cern.ch/lhcathome/forum_thread.php?id=4611#34323