... but is it not strange that i've been running T4T@H for a whole month now along side E@H without any E@H cache problems until very recently?
No, it's not really that strange because until the debt to the under-supplying project becomes large enough, there tends not to be the high level of restriction on the project that can always supply. I would guess that it probably took a while to get to this critical level of debt where E@H will only be asked for work as a last resort when the cache is about to run totally dry. BOINC dosen't really handle well, projects that often can't supply work when requested.
Quote:
by the way, my cache seems to be all over draining again...
Not surprising, since the large debt to the other project is still there. As a temporasry workaround, you could try zeroing debts manually. As a more permanent workaround, you could try adjusting resource shares to more accurately reflect what each project is able to deliver. So, if the under-supplying project can only supply work about 10% of the time, give it a resource share of only 10%. That way a large debt to the under-supplying project shouldn't be able to build up. When the under-supplying project can supply more, give it a bigger resource share. By doing that, you should always be able to keep your E@H cache full.
Please be aware that how BOINC handles resource shares and debt has changed at various stages over the history of BOINC. The behaviour I describe above was certainly evident in older versions and might be different in more recent versions. I have no experience with the latest versions and you may need to do a bit of experimenting in order to find the solution to your problem. You may also get better behaviour by not having an overly large cache size.
thanks Gary...that more or less makes perfect sense. i didn't think about the fact that i shouldn't expect to see anything out of the ordinary until a sufficient amount of debt has accumulated. and while E@H typically downloads new tasks on a regular and frequent basis, T4T@H only downloads one new task every 24 hours. for those who aren't familiar with the project, T4T@H is actually the LHC@Home 2.0 Beta project. but unlike it predecessor, tasks for LHC@H 2.0 Beta are actually crunched in Oracle's VirtualBox virtual machine using CERN's (European Council for Nuclear Research) custom version of Linux. a wrapper is implemented to allow the project to run in BOINC under the project name Test4Theory@Home and earn credit. while the virtual machine may very well crunch several CERN tasks per 24 hour period, the associated T4T@H task in BOINC will actually run for a full 24 hours before completing, at which point it will upload and a new T4T@H task will download and start.
so if its expected that T4T@H will only fetch new work once every 24 hours, i just need to calculate the appropriate allocation of resources between projects. i'll play around with resource allocation and report back...
ok, i've since adjusted my resource allocation between projects to more accurately reflect what they should truly consume. i also played with my cache size, so it'll probably be a few days before a burn through a decent chunk of the existing cache. in other words, it'll be a few days before i know if E@H starts fetching work on a regular basis given the changes i've made, or if its going to wait until my cache is near empty before downloading any new E@H work. i'll report back when i know more...
RE: ... but is it not
)
No, it's not really that strange because until the debt to the under-supplying project becomes large enough, there tends not to be the high level of restriction on the project that can always supply. I would guess that it probably took a while to get to this critical level of debt where E@H will only be asked for work as a last resort when the cache is about to run totally dry. BOINC dosen't really handle well, projects that often can't supply work when requested.
Not surprising, since the large debt to the other project is still there. As a temporasry workaround, you could try zeroing debts manually. As a more permanent workaround, you could try adjusting resource shares to more accurately reflect what each project is able to deliver. So, if the under-supplying project can only supply work about 10% of the time, give it a resource share of only 10%. That way a large debt to the under-supplying project shouldn't be able to build up. When the under-supplying project can supply more, give it a bigger resource share. By doing that, you should always be able to keep your E@H cache full.
Please be aware that how BOINC handles resource shares and debt has changed at various stages over the history of BOINC. The behaviour I describe above was certainly evident in older versions and might be different in more recent versions. I have no experience with the latest versions and you may need to do a bit of experimenting in order to find the solution to your problem. You may also get better behaviour by not having an overly large cache size.
Cheers,
Gary.
thanks Gary...that more or
)
thanks Gary...that more or less makes perfect sense. i didn't think about the fact that i shouldn't expect to see anything out of the ordinary until a sufficient amount of debt has accumulated. and while E@H typically downloads new tasks on a regular and frequent basis, T4T@H only downloads one new task every 24 hours. for those who aren't familiar with the project, T4T@H is actually the LHC@Home 2.0 Beta project. but unlike it predecessor, tasks for LHC@H 2.0 Beta are actually crunched in Oracle's VirtualBox virtual machine using CERN's (European Council for Nuclear Research) custom version of Linux. a wrapper is implemented to allow the project to run in BOINC under the project name Test4Theory@Home and earn credit. while the virtual machine may very well crunch several CERN tasks per 24 hour period, the associated T4T@H task in BOINC will actually run for a full 24 hours before completing, at which point it will upload and a new T4T@H task will download and start.
so if its expected that T4T@H will only fetch new work once every 24 hours, i just need to calculate the appropriate allocation of resources between projects. i'll play around with resource allocation and report back...
ok, i've since adjusted my
)
ok, i've since adjusted my resource allocation between projects to more accurately reflect what they should truly consume. i also played with my cache size, so it'll probably be a few days before a burn through a decent chunk of the existing cache. in other words, it'll be a few days before i know if E@H starts fetching work on a regular basis given the changes i've made, or if its going to wait until my cache is near empty before downloading any new E@H work. i'll report back when i know more...
Don't forget not to install
)
Don't forget not to install BOINC 6.13.1 or higher. Only 6.12.34. Or you can drop into the problem I recently had.