I stand to be corrected, here, but it would appear to me that if the "additional work" cache is smaller than the estimated run-time for a aingle AP then you will never be allocated an AP? After all, the work request actually requests xx seconds of work so I would not expect a task with an estimate of 10 * xx to be allocated - or does it just allocate the next task and then think "Oops - no more space left"?

It doesn't matter if cache-size is one second or larger, you'll still need a full wu to fill this cache-size.

Afterall, if work-cache needed to be larger than wu-size, no-one would download any of the 3-month+ CPDN-models. :)

Dohhh... How did that pass me by? I have a couple of those on the go!!

Let's put it this way then. A 0.1 day cache equates to just under 2.5 hour's work so a machine with such a setting will never ask for more than 2.5 hours worth of work. If that box is a P4 then the first WU allocated will satisfy the request, whether that task is MB or AP i.e. if you want AP then the next task to be allocated has to be an AP. On the other hand, when my quaddie tops up its 3-day cache it often pulls down 20 or 40 tasks in a short time so for me to get AP it doesn't matter if then next task is AP as long as there is one (or more) AP in the batch that I get. Seems to me that I have a greater chance of snagging AP work? Obviously the same goes (in spades) for the top crunchers.

Now, someone with a PhD in probability theory come and shoot me down :)

F.

---

On a related topic, I have seen many questions along the lines of "are we going to have the ability to select AP5v5 on the preferences page?" but I've never seen any form of satisfactory answer. By that I mean "no", not really satisfying but at least clear; or "yes in xx days".

I don't know, but it seems intuitively obvious why we haven't seen an official answer.

They haven't told us "yes, 3 days from now" because they have more important things to do -- and they're things that we'd rather have them do, like get a stronger, more stable replica database, work out upload/download issues, etc.

... and because, if they say 3 days, and something happens that is more important, they'll get complaints.

Matt said that he's been spending most of his time doing systems stuff and not working on writing code -- and I expect it's the same for Eric and Jeff.

I'm also guessing that there will only be a few minutes between "okay, let's do this" and when it is done. I don't expect an announcement at all, because someone will notice in a few minutes.

---

I have run a cache most of the time at ,10 and did a lot of the old AP and a few of the ver 5,03 .
I would not think that the size of a cache would limit getting an AP.
I run milkyway also, If my seti cache is to big i cant get work units from them because i have 6.6.36. Im waiting for 6.6.37 to be officially released.
maybe its a 6.6.36 issue.

---

[/quote]

Old James

I'm also guessing that there will only be a few minutes between "okay, let's do this" and when it is done. I don't expect an announcement at all, because someone will notice in a few minutes.

Ned, I'm not expecting an announcement, just asking if anyone knows what's planned. Don't shoot me :)

Hi, Matt, Eric, Jeff, et al: The upload server is not working! (It's been down since about 2:30 AM Pacific Time)

Please don't go home for the weekend without first kicking the upload server! Thanks!

Plaese anyone give the upload server a kick start.
Over 200 wu's trying to upload for the last hour or so.
Thank's in antiticipation

---

Now there is realy something odd going on. After some managing I was lucky to get an AP along with two short MB's. So far so good. But a few minutes ago I looked to my resultspage again and one of the older results waiting for validation changed completely, was about 3000 secs and look at the time the wingman suddenly got to complete:

http://setiathome.berkeley.edu/workunit.php?wuid=482671080

I'm also guessing that there will only be a few minutes between "okay, let's do this" and when it is done. I don't expect an announcement at all, because someone will notice in a few minutes.

Ned, I'm not expecting an announcement, just asking if anyone knows what's planned. Don't shoot me :)

I'm not armed.

I am suggesting that when a question has been asked several times, and gone unanswered, it seems IOTTMCO that asking again will get the same result.

---

I wonder if a schedule could be developed -- we know of the regular 4 to 6 hour Tuesday outage and 2x post outage congestion, but it seems there is something of a regular Friday/Saturday upload server outage (planned perhaps) and post upload server outage congestion as well. Then there is the download congestion when AP workunit batches are released.

Might be of use to have this information.

---

I wonder if a schedule could be developed -- we know of the regular 4 to 6 hour Tuesday outage and 2x post outage congestion, but it seems there is something of a regular Friday/Saturday upload server outage (planned perhaps) and post upload server outage congestion as well. Then there is the download congestion when AP workunit batches are released.

Might be of use to have this information.

Assuming, of course that scheduled outages are the only kind.

---

Grumble, I knew I should have gotten more WU last night... Hopefully it's not down for more than (another) day.

Now there is realy something odd going on. After some managing I was lucky to get an AP along with two short MB's. So far so good. But a few minutes ago I looked to my resultspage again and one of the older results waiting for validation changed completely, was about 3000 secs and look at the time the wingman suddenly got to complete:

http://setiathome.berkeley.edu/workunit.php?wuid=482671080

But the wingman returned a -9 (exited early). This can be caused by a fault in the PC so is not necessarily indicative of a problem with the WU.

F.

---

1313775231 4616851 22 Jul 2009 21:32:01 UTC 24 Jul 2009 18:10:47 UTC Completed, waiting for validation 787.67 13.16 pending
1313775232 4134823 22 Jul 2009 21:32:03 UTC 7 Sep 2009 2:48:08 UTC In progress --- --- ---

strange don't you think?

1313775231 4616851 22 Jul 2009 21:32:01 UTC 24 Jul 2009 18:10:47 UTC Completed, waiting for validation 787.67 13.16 pending
1313775232 4134823 22 Jul 2009 21:32:03 UTC 7 Sep 2009 2:48:08 UTC In progress --- --- ---

strange don't you think?

But check the Report Deadline in the wingman's returned result:

Report deadline 7 Sep 2009 2:48:06 UTC

F.

---

Maybe - to avoid confusion in the future - there should be a list made available with all regularly scheduled outages, reasons for that and estimated downtimes.

Even if it's a non-scheduled outage a note in the tech news would be nice to know what's going on and how long servers will be down.

[quote]Now there is realy something odd going on. After some managing I was lucky to get an AP along with two short MB's. So far so good. But a few minutes ago I looked to my resultspage again and one of the older results waiting for validation changed completely, was about 3000 secs and look at the time the wingman suddenly got to complete:

http://setiathome.berkeley.edu/workunit.php?wuid=482671080

Linked it for you.

---

[/quote]

Old James

Wich makes it more odd cos that is me and it was reported july 24

http://setiathome.berkeley.edu/results.php?hostid=4616851

Besides non so short a wu will get that long time to complete.
And there is that change in time completed as I mentioned. But I'm going to sleep.Will read tomorrow more about this phenomenon.

Besides non so short a wu will get that long time to complete.
And there is that change in time completed as I mentioned. But I'm going to sleep.Will read tomorrow more about this phenomenon.

OK, I missed the fact that your machine returned the -9. But this is a standard 0.4x Angle Range WU so it doesn't warrant a short turn-round time - and it didn't have one as the line in my previous post is the target date from your host. So far as I can see, your host got to the limit of 30 spikes/etc. and bailed out early. I can't see any way that it can have shown previously on the Tasks page as taking ~3000 sec.

F.

---

Not at all, scheduled outages are one (and for SETI) significant set of outages. Knowing when they are would (in my opinion) be useful.

Of course there are unscheduled outages -- this is life, in fact, this is life on SETI. To think I was assuming that all outages here are scheduled would be suggesting I haven't watched the process here over the years -- and I would assume you actually know better than that.

Assuming, of course that scheduled outages are the only kind.

---