General Preferences - I need some guidelines

One of my computers is a 500 mhz box. It usually completes a WU in the 140,000 to 160,000 CPU-sec range. That's a little less than two days of CPU time. Your box will about take about 10% longer but you should be able to meet the 2 week deadline as long as you keep Einstein's time percentage high enough to do 2 days work every two weeks (e.g. If your computer is on all the time, 20% or higher should do; but, if it's only on 8 hours per day, then you might need 60%.)

I am new to BOINC and to Einstein. I'm running a 700 MHz Mac G4 with 512 MB of RAM and I get through an Einstein WU in ~54,200 seconds (a shade over 15 hours CPU time). I have four questions if I may about general preferences and about Einstein in particular.

(1) How often does Einstein lay down checkpoints?
(2) Can you force Einstein to make an unscheduled checkpoint if you have to shut down the computer for any reason and thereby avoid losing work?
(3) What is the meaning of the General Preference option "write to disc at most every 60 seconds"? Does it have any relationship to the checkpoint interval? The reason for asking is that I would like to keep the amount of disc activity to a reasonable minimum, consistent with the proper operation of the client.
(4) What happens if for any reason the client can't make a scheduled contact with the server? Does it keep trying, or does it wait until the next contact is due?

Any advice would be appreciated.

I think I know what Jim meant: If you are using one of the newer BOINC's, the program will adjust its work schedule, in an attempt to meet all your deadlines. For example, if an Einstein unit deadline is approaching and your normal resource share does not give enough time to finish it, BOINC will "borrow" time from SETI in order to meet the deadline. Then, when it finishes the Einstein unit, it will "pay SETI back" for the time it "borrowed". When, the "pay back" is in process, BOINC may not ask Einstein for new work. Hence, "you will not always have work from all projects on your computer at the same time".

One more thing: Since you are new to Einstein, you may not know about this issue yet. If you are using BOINC 4.45 and Einstein 4.79 (the Windows versions), and if you are using the BOINC screensaver (or "Show graphics" option), you may encounter the "graphics bug" - a problem that ends WU processing early with "Client error Computing" messages. If this happens to you, try the Beta application (under More Information on the E@H home page).

99.94,

I'm not a MAC guy either, but it should work the same across platforms.

Q1-3: The "write to disk at most every __ seconds" is the checkpoint interval. AFAIK, you can't "force" a checkpoint. The default is 60 seconds, so it writes the checkpoint every minute. Check the file Fstats.ha.ckp and fstats.hb.ckp in the slots/n folder.

If you set the write interval to 600 seconds (10 minutes) you can lose that much crunch time if the application exits. [EDIT] It doesn't write a checkpoint when it exits so you can lose whatever was done since the last checkpoint. Checkpoints don't actually do that much, BOINC writes a new client_state.xml file and the Einstein application writes a pointer to the fstats.h*.ckp file. Its not like it writes the contents of memory each time.

All this is done on the host, nothing gets sent to the project servers until the application completes the workunit, then BOINC sends the result to the server.

Q4: Are you asking about BOINC to project server communications? If it can't connect it'll wait and retry. The wait time increases so it doesn't constantly retry if you have network problems.

If you're asking about the application to BOINC communications, the app checks for "heartbeat" messages from BOINC, if it doesn't see any for 30 seconds it exits. If this happens you'll see a "no heartbeat" message in stderr.txt - look in the file (slots/n) while its running, or in the result when its complete, uploaded and reported. If BOINC is running you'll see a "app exited with zero status but no finished file" message. That happens occasionally when something ties up BOINC for >30 seconds.

Walt