The ARP floodgate opened. I just got a batch of ARPs (3)

The reliable status has to do with a certain amount of valid ARP WUs returned on time.

Regarding "reliable status" and ARP1 task acquisition in general...

Whilst I think "Extreme" tasks only get sent to so-called reliable machines, I'm not convinced that the current general oddities in task acquisition are related to that status. I suspect that new tasks only get added to the mix at certain times of day, and if a client doesn't happen to ask for work whilst that work is available it will only be likely to pick up the occasional retry (which should not be affected by any work-flow controls that might be in place!)

As for "reliable" in general, according to the documentation a reliable host needs to have submitted a sequence of at least 10 consecutive valid results. If tasks are marked as needing a reliable host it may also take into account the average turnaround time for the host (though I'm not sure how well that aspect works!)...

My four systems that accept ARP1 work only have non-valid results when there are upload/download problems, and all have small enough buffers to guarantee that any standard 64-bit run will be returned within 24 hours and have an execution time under 12 hours; despite that, some of my systems still end up short of work if the BOINC client doesn't request work regularly. That can't be down to "reliability" :-)

Cheers - Al.

P.S. I suspect there are still users who run click-scripts to ask for work every 5 minutes or so -- in the past several forum posters have described doing something like that to keep their buffers topped up... I wonder if they also have problems getting work? [I'm not recommending requests that frequent, by the way :-), though I do wish the client wasn't as quick to increase the request wait time when it doesn't get work!]

Yes, lots of "Tasks are committed to other platforms", before any tasks are sent.

Hi Unixchick,
Even with a short queue it seems difficult - in my experience - to keep a full queue.
Example: I'm trying to get 16 ARP1s in my shorter queue of 64 MCM1s and I've only once managed to get a full queue of 16 ARP1s in the past week.
During the past 72 hours, hour by hour, I've measured the number of ARP1s (In Progress) on that host and this is the result:

 8  7  6  6  7  7  7 10 13 15 15 16 15 14 13 10 12 13 15 15 15 13 12 11
11 10 10  9  7  8  6  7 10  9  8  8 10 10 10 10 11 11  9  8  6  6  6  7
 8  8  8  6  6  6  5  5  5  5  5  6  7  7  7  7  9 10  9 10  9  9  8  6 <-- most recent measurement

Adri

Interesting data Adri. I'd really like to see them send out more ARP. It really shows how variable the amount of WUs is. It doesn't seem to be on an automated schedule.

I haven't seen an ARP in a while, and MCM don't always come on the first ask. The system is limping through the weekend.