linux 4.21 doesn't validate against windows 4.17?
4 Jun 2007 9:23:05 UTC
Topic 192810
(moderation:
Over the last few weeks, I've had a number of results which didn't validate.
In every case, this has been where my host (linux, running 4.21) was validating against windows hosts, running 4.17; the windows hosts seem to have won. I have never seen a problem where my host is validating against other linux hosts, my results are always valid otherwise.
An example of such a workunit is http://einsteinathome.org/workunit/33809133
Is there some sort of cross-platform validation issue going on, or should I suspect my host?
linux 4.21 doesn't validate against windows 4.17?
)
There is a known cross-platform validation issue as you suspected, no need to mistrust your host :-)
CU
BRM
Oh... and I see could have
)
Oh... and I see could have found that out myself if I'd looked a bit harder in these forums. Doh. Thanks very much for the reply, BRM.
Here's another one that lost
)
Here's another one that lost out in the Linux (me) versus Windows validation problem. I have been lucky so far as this is the first one I seem to recall, but with about a dozen still pending I could get others.
33948225
Hope it helps.
Here is a result where I lost
)
Here is a result where I lost out, with me running Windows app 4.17. The other two results validated to make quorum, were obtained from Linux app 4.21.
http://einsteinathome.org/workunit/33926687
We should be getting credit for these "invalid" results, since the problem appears to be with the Linux app 4.21 beta.
LP.
I've had 3-4 linux units
)
I've had 3-4 linux units thrown out with same problem in last few months. Most recent will be 33926016 (http://einsteinathome.org/workunit/33926016). You can tell because as soon as you report result it is sent to a third machine. Actually the problem goes both ways. The subsequent results have to match the first machine to complete successfully. I had one result where my linux machine finished first and a windows machine received no credit as 2nd and the unit was sent to 3rd machine which was running linux.
This seems to affect linux machines more because I believe the linux machines are generally older and slower and more heavily loaded. The linux machines will thus finish second more frequently. This is true in my case at least.
Admittedly my coding days are long passed, but it does not seem to me it would be that difficult to program E@H servers to check operating system of clients and only to send a particular WU to machines with same operating system. I don't know if this is a boinc problem or an E@H problem since I only run E@H. It does make you wonder if it is worthwhile in terms of cost to client and project to run E@H on these older linux machines.
RE: I've had 3-4 linux
)
Actually the ability to check host platforms is already built into the BOINC framework. It's called Homogenous Rendundancy, but the issue is it was never required on EAH before. The problem has been with the new app being tried out now on this defacto beta S5R2 run.
Other than that EAH is one of the most old host friendly projects around, since the don't issue a useless result by default just to keep the the DIG's happy (Demands Instant Gratification). This means when things are 'normal' around here you are reasonably assured that if your host completes the result successfully and on time, the effort it put in was actually helping the science.
Alinator