Large downloads
30 Mar 2008 13:15:21 UTC
Topic 193595
(moderation:
I have been running Einstien@home for about 3 years on 4 machines and periodically I would of course have to download new work usually in the form of 3.08 and 3.38 MB files. For some reason on a machine running a 2.07 GB 64 bit Athlon that normally finishes a job in about 14-15 hours I was only having to download new work every couple of weeks, sometimes it would be in the form of one 3.08mb file and one 3.38mb file. Other times it might consist of 5 or 6 files of each size.
However over the past 3 days when a unit is approx. 90% completed, where before it would bring up a new job called "ready to run", now I am getting a huge download consisting of as many as 20 3.08mb and 3.38mb files. This occurs every time 1 single 14 hr job is completed. In other words, this machine is spending far more time downloading that crunching data. Anyone have an idea what the problem might be. I have change nothing on my preferences for several months.
F. Prefect
In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.....Douglas Adams
Large downloads
)
Hi!
I've no explanation for this particular problem, but in general you might want to increase your WU "cache" in the web preferences so that you are (almost) guaranteed to get more work for the same set of input data files.
Still this seems strange. Are the downloads succeeding or could those frequent downloads be retries for the same downloads?
Bikeman
I believe this is just a
)
I believe this is just a somewhat unfortunate characteristic of how the work scheduler behaves. At the moment we seem to be having a mini-cleanup of odd outstanding tasks as we shift to new higher frequency work.
You will recall that a little while ago (maybe 6-8 weeks) we started on >800 frequency work. In a fairly short space of time I saw on my machines skygrid files for frequencies ranging from 800 to about 930. Until quite recently, any given machine could work through many tasks without having to download new large data files - exactly the files the OP is referring to. Ocassionally, a couple of the lower frequency steps might be marked for deletion and a couple of new large files at the higher end of the frequency steps would be added.
That period is drawing to a close and I've seen quite a few examples now of machines getting new skygrid files in the 940 to 990 range, as tasks for the previous lower frequencies become exhausted. Before transitioning to the much higher skygrid frequency, a host may well be given some odd tasks to cleanup from elsewhere in the frequency range covered by the former skygrid file. Quite often these may be scattered sufficiently far apart in frequency (but still within the same skygrid range) to require a lot of new data files for just one or two tasks. I have seen this several times myself and there is an example of it in this results list for one of the OP's machines. Currently there are 15 tasks visible and here is a list of their frequencies and sequence numbers.
Hopefully by looking at the above list you will see what I'm trying to say :-).
Now that the host has transitioned to the new 990 SkyGrid range, there will be quite a lot of tasks in the sequence that will not require further downloading of large data files for quite some time. As other hosts are given these same frequency steps, the OP should see some increase in the gap between seq#s that his host is issued. When these gaps start to become very big it will be a warning that this frequency range is nearly exhausted as well.
Cheers,
Gary.
RE: In other words, this
)
Just a small correction here. Your machine continues to crunch the current task while new data is being downloaded. The actual downloading of data probably wouldn't take that much in the way of resources so your machine wouldn't be spending more time downloading than crunching. Unless your downloading speed is so slow that you actually run out of work on hand, I would guess that you would lose very little in crunching time. The bandwidth consumed is far more annoying, particularly if there are lots of hosts involved.
Cheers,
Gary.
Could it be he's concerned
)
Could it be he's concerned with the DL time because he is connected to dialup...I know it was frustrating for me to have to wait for large files to transfer back in the days when i was using dialup...20mb would take hours.