The only way I am keeping something going all the time is running both a Windows and a Linux VM in my Linux host. Even with three clients running on the physical machine, I am often down to just one task running.

The supply of ARP1 WUs this morning is less than half of what it was in the last few days, is there a known problem or just something brewing up ready for the coming weekend?

Ralf

I have 8 threads on each machine and use app_config to restrict ARP1 to 4 threads on each machine. That way if either has less than 4 ARP1 units, it will fill up with MCM1.

As MCM1 units complete in about a tenth of ARP1, they request new work 10 times as often including any ARP1 that may be available.

Mike

Mike,

Looks like a reasonable [and feasible] summary; to me, the interesting part would be the fine detail of the stages between work-unit assimilation and new work-unit number allocation, and whether some sort of [standard] BOINC priority tagging is used to give precedence to tasks independent of WU number order!

It would indeed be nice to know the entire process, though I do wonder if the WCG Tech Team is still at the point of "We just use what IBM provided" for some of the bits and pieces that make use of non-BOINC mechanisms, in which case we are unlikely to get any enlightenment :-(

By the way, regarding the 448 units that don't show up in the state.txt count: there were only two missing units when the system went into the "no storage" 2023-2024 hiatus, but when they started work on getting ARP1 restarted after that the count went to 446 almost at once [2024-07-18] (presumably because of whatever WCG did to get things ready).

When work actually started to flow again [2024-10-31?] there was a period of a week where Normal work-units completing were rapidly re-categorized as Extreme or Accelerated, not really surprising given what appeared to me to be a pre-hiatus state.txt imbalance toward Normal (implying some units hadn't run for a long time). Within a week [2024-11-04] a further two WUs "disappeared" and the count has been stuck at 448 ever since!

It would be interesting to know whether those items are all [deliberately?] blocked or whether they have dropped into a data hole that has something to do with how the ARP1 restart happened...

[Edited to add the below]
There was a similar (but much larger) mass disappearance of 6423 units from state.txt on 2022-11-01 -- apart from a very small change that appeared to coincide with them restarting the three ultras early in 2023, things remained unchanged until 2023-05-19 when suddenly there were only 2 missing from state.txt and that stayed the same until the next hiatus (as mentioned above). According to the Project Statistics there was a massive increase in work returned around that time, so some sort of blockage was definitely dealt with, presumably about a week earlier given the deadline on Normal tasks...

Again, was the blockage because all those units had genuinely ended up in an Error state or was it a non-BOINC database issue (or was it a mixture of the two!)?

Cheers - Al.

Sunday Report

ARP seems to be running smoothly but with only a few units coming out. I haven't seen any for the last 30 hours.

This report has been rebased since the January restart. Any forecasts will be based on average throughput for the weeks availailable until 5 weeks are available whence it will be based.on the last 5 weeks. Next week will be the 5 week.

All classifications have stayed put..

There are still 3 ultras in generations 21 & 22.

There are 18 units seemingly stuck in generations 104 to 121.

There are 61 other extremes missing from state.txt presumed stuck.

There are now 337 extremes in generations 115 to 131.

There are now 550 accelerated units in generations 132 to 136 of which 417 are missing,presumed stuck.

There are 34,719 normal units in generations 137 to 146 with 51 extras listed .

The highest generation to have had validations is still 141.

20,810 units have validated this week.

Based on the 4 weeks, we would complete ARP1 in late 2026 but there are fewer units currently being issued.

Mike