At the time I saw this I was helping out another guy get his CFSW( I think) badge, short WU but LOTS of memory usage. Now back on my account doing a mix of projects and using 5.6 gig of ram memory.
Single 128 gig Crucial M4 SSD on this machine showing 21.5 gig free on the drive.
I still think that SOME sort of limitation should be put into place for the priority WU..How about a max of "1/2 available cores" or threads and that would keep it in check for even the people with limited ram memory..
Ram permission is set to 100% BUT I just checked and set for 5 day cache not 3, just changed it down. That may have been part of the issue I saw.
Who's Murphy?
And a tease: People still using those "old" Octo machines?
I'm out looking for a pair of the upcoming 15 core IB xeons..30 cores/60 threads,pops right into this board and only 130W vs the E5-2687W 150W..
Thanks again for the info..Appreciated.

Boinc does NOT always run tasks in return by date order, it runs tasks based on the priority assigned by the format the programmers put in it. As you get closer to the return by date Boinc WILL get more into a 'logical' to us humans order, but that is NOT the normal way Boinc works.

As for flooding you with work, you COULD upgrade the memory to 32 or even 64 gb but that is your choice just as it is to keep the units in memory when suspended. YES there is good reason to do this but not if it is causing other problems. As far as only sending one unit to a 'trusted' machine, that would seem to be hoping that there are TONS of other 'trusted' machines and that may simply not be the case. 'Trusted' machines are usually machines that can be 'trusted' to return a unit in a short amount of time AND not have any problems with the units. YES obviously there is more than one 'trusted' machine here at WCG, we ALL like to think our machines are in that category, but the plain fact is that they are not ALL 'trusted'. Sometimes there are more units needed to be resent than other times, you seem to have gotten a large stack of them. Do you crunch for ALL projects here at WCG, or just some? Obviously if only some that will reduce the amount of units you can potentially get. BUT it may also increase the number of units you get as your pc is in the 'trusted' group and there may not be that many others in those projects.

There are just soooo many variables that simple answers are just not possible, even though we would LOVE to have them. It is kind of like asking someone to point to the ocean, no matter which way they point there is an ocean in that direction! Asking them to point to the CLOSEST ocean results in a much clearer answer, but is that as the crow flies or by following the road? There is a Boinc Mailing List run by the Boinc Programers at Berkeley and specifically by Dr. David Anderson, the original writer of Boinc. It is open to anyone but be aware that Boinc is Dr. Anderson's baby and he does NOT take criticism well. There are over half a dozen programmers helping him keep Boinc updated, but Dr. Anderson is the main man. I have personally always found Senior Programmer Rom Walton to be VERY open and helpful, but even he sometimes does what he is told as he is not his own boss. Boinc was first created, for the public, in 1999 and it has gone thru MANY changes, for MANY different reasons, to get to where it is today. The mailing list can be helpful in seeing what is going on and in also seeing some of the why things are the way they are.

For the most part, I have observed BOINC does run in FIFO order, except when a repair task or beta unit is present. At such time it may suspend a running task and start immediately or it may wait until the first of the running tasks is finished and then start. As a rule these tasks always have shorter deadlines. However, I have noticed on my one octo machine running Linux Mint 12 (6.10.58)with a 3 day cache setting some different behavior, which only seems to be occurring as of the last month or so. Because I only upload/download every three days big chunks of the available work units all have the same return date with time only varying by a few minutes. Randomly this machine will suspend all eight running units at various stages of completion and start other units with the same date/time. These run in high priorty mode. Once it finishes these units it gradually will return to the suspended units and finish them. There is never any danger these units (or the entire cache) will not finish in time. Two things come to mind on why this may occur. The work units run into a few which turn out to run longer than the original estimates so BOINC readjusts upward how long the rest of the awaiting units will be expected to run so it starts some in high priority mode. If these run shorter, it readjusts again to lower the expected run run time of the rest of the cache and goes back to normal mode. The variability of the run time of the work units may be the culprit. The project is HFCC run on this machine with 150 to 200 tasks in the queue.
The second item may be what mickey159b states: There is something embedded in the work unit(s) which specifies a time which BOINC algorithms read which causes the hurry up behavior.
At any rate I just leave it alone and it gets everything done in a timely fashion.
Cheers

You are correct in that I have never seen a repair unit on this machine.
Cheers

THANK YOU....for the life of me I could NOT remember FIFO when typing all that!!! I am getting frickin old!!!!