... Either the solutions/theories being presented are way over my head, or you guys don't know s..t.

Likely quite a bit of both! (Sorry for the half-bit pun...)

Having been through several major SETI breakdowns in the last few years, I have learned patience, and to have other projects on the go. My ultimate fallback at the moment is climateprediction.net. ...

And that is likely the best strategy.


The outbound choking will also choke the ACK packets needed for the inbound data. If there is no QoS in the routers on each end of the choked link, then any data packet can get randomly dropped and lost, including the all important control packets. Hence, the inbound 'appears' to be also choked even though physically it isn't.

Unfortunately, junking the control packets is very expensive in lost bandwidth because you waste whatever associated big data packets were associated with each lost (tiny) control packet.

A combination of nefarious QoS using delay buffers rather than 'policing' and tweaks to the Boinc protocols could well smooth out the surge to something more graceful.

Then again, the present system blunders along well enough and there's more urgent fixes to be fixed first.

In any case, if the system was perfect, we'd have nothing to grouch about!


Happy crunchin',
Martin

[edit]
Actually, if you assume a limit of something like 95% of the link's physical limit, you can do clever tricks with QoS on just one end of a choked link. Even if their router box doesn't have that facility, a medium spec diskless linux box can do the job nicely just from what is already available in the kernel...
[/edit]

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

Actually, if you assume a limit of something like 95% of the link's physical limit, you can do clever tricks with QoS on just one end of a choked link. Even if their router box doesn't have that facility, a medium spec diskless linux box can do the job nicely just from what is already available in the kernel...

.... and all of this overlooks one really cool fact that BOINC should take advantage of, and does not.

BOINC is a system. It is a BOINC client talking to a BOINC server.

It is possible (with changes to the client, of course) for a project to "throttle" the clients when the load is high -- if the load is over 90%, tell the clients to back off a bit -- or a lot.

If the project is down, it should be able to tell the clients to just stop connecting for a while.

When the project is back, holding the demand to 80% of capacity will let each connection finish without wasting bandwidth on failed connections, and CPU cycles on excessive SYN processing.

Of course, there is an issue getting out the word if the project servers are heavily loaded. The answer is left as an exercise to the reader.

---

Someone on staff should check the stats .xml export - I think it may be down...
Hard to tell with all the ups/downs, but I haven't had a SETI credit since Sunday. (and I've uploaded about 80 Wu's in that period!)

STILL the case...

---

.

Hello, from Albany, CA!...

The following WU's are ghost, they do not show up anywhere in the client:

949023054 313897646 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023053 313897640 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023051 313897634 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023049 313897628 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023046 313897622 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023044 313897616 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023042 313897610 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023040 313897604 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023038 313897598 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023036 313897592 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023034 313897586 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023032 313897580 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949023030 313897574 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949022964 313897510 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949022961 313897498 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949022958 313897492 13 Aug 2008 0:27:24 UTC 20 Aug 2008 0:27:24 UTC In progress --- New --- --- ---
949022957 313897486 13 Aug 2008 0:27:24 UTC 30 Aug 2008 7:45:42 UTC In progress --- New --- --- ---

There was a timeout on a work fetch scheduler request within a few seconds of the time that they were supposed to be sent.

those show as "new" - they haven't been assigned to a client yet!

---

.

Hello, from Albany, CA!...

Someone on staff should check the stats .xml export - I think it may be down...
Hard to tell with all the ups/downs, but I haven't had a SETI credit since Sunday. (and I've uploaded about 80 Wu's in that period!)

STILL the case...

Same here, my three machines are putting up numbers, but no pending credits have been released for 6 days. This seems to be longer than usual.

---

[quote]Someone on staff should check the stats .xml export - I think it may be down...
Hard to tell with all the ups/downs, but I haven't had a SETI credit since Sunday. (and I've uploaded about 80 Wu's in that period!)

STILL the case...

Same here, Since the big down time,I'm not posting any credits, and what really hurts is some are from Astropulse.... they really count. I've also noticed, although this could be on my end, some Compute errors, which I've never had. And yes.... one was for s huge Astropulse job. Even sadder is that on some of my slower machines, they are justr about to pass their due date, because thet've been sitting almost a week.

Anyone have any good news to say we'll eventually get credit - at least for the in progress?

---

I'm seeing my stats change here on the website and seeing my pending numbers go down. I'm not that worried that the xml exports haven't made it out yet. Just think of how the numbers will look once it happens.

Remember, Matt's been out on vacation this week so they are one man down. I would expect to see it addressed early next week. Interestingly enough, the Seti Beta sent out stats yesterday I believe....

---

Wendy



Click Here for BOINC FAQ Service

Same here, Since the big down time,I'm not posting any credits, and what really hurts is some are from Astropulse.... they really count. I've also noticed, although this could be on my end, some Compute errors, which I've never had. And yes.... one was for s huge Astropulse job. Even sadder is that on some of my slower machines, they are justr about to pass their due date, because thet've been sitting almost a week.

Anyone have any good news to say we'll eventually get credit - at least for the in progress?

I wouldn't worry too much about your WUs about to pass their dates - if there are problems at the SETI end, these WUs are usually honoured.

You'll get the credit - eventually!

---

[quote]


I wouldn't worry too much about your WUs about to pass their dates - if there are problems at the SETI end, these WUs are usually honoured.

You'll get the credit - eventually!

Thank you, that's encouraging....I've just never flatlined for so long

---

I'm seeing my stats change here on the website and seeing my pending numbers go down. I'm not that worried that the xml exports haven't made it out yet. Just think of how the numbers will look once it happens.

You're lucky than. My pending credits are rising, my average credit is still dropping and my stats at boincstat are on hold for 5 days now :)

---

-=[ Not all who wander are lost ]=-

I'm seeing my stats change here on the website and seeing my pending numbers go down. I'm not that worried that the xml exports haven't made it out yet. Just think of how the numbers will look once it happens.

You're lucky than. My pending credits are rising, my average credit is still dropping and my stats at boincstat are on hold for 5 days now :)

It's encouraging to learn we are not alone (no pun intended). I guess there's no choice but to wait and see. The good news is, I'll find out if I was able to correct my signature.
Keep on crunching!

---

[quote]

snip

[quote]

snip

It's encouraging to learn we are not alone (no pun intended). I guess there's no choice but to wait and see. The good news is, I'll find out if I was able to correct my signature.
Keep on crunching!

[img]http://www.boincstats.com/signature/user_251558.gif[img] <---- forgot the slash ---->[/img]

---

BOINC Wiki . . .

Science Status Page . . .

Thank you. It only took 9 years for me to get her.

---

I'm really starting to get worried now. It's been a week (since the event) that I have received almost no credit. Everything here on this end appears normal. I see WUs returned the 12th and 13th but only a fraction of what was crunched. I am aslo seeing many as pending, or still showing as new, with no entries for the 14th and 15th, and a few returns for 16th and 17th. Even taking into consideration that I am returniong fewer while running Astropulse, it just doesn't feel right, especially since I've returned a number of AP WUs.

Since I am using multiple computers, I must assume my end is good. Am I just being paranoid, or can I expect to see a HUGE credit on Monday when everyone is back in the helm?

---

I'm really starting to get worried now. It's been a week (since the event) that I have received almost no credit. Everything here on this end appears normal. I see WUs returned the 12th and 13th but only a fraction of what was crunched. I am aslo seeing many as pending, or still showing as new, with no entries for the 14th and 15th, and a few returns for 16th and 17th. Even taking into consideration that I am returniong fewer while running Astropulse, it just doesn't feel right, especially since I've returned a number of AP WUs.

Since I am using multiple computers, I must assume my end is good. Am I just being paranoid, or can I expect to see a HUGE credit on Monday when everyone is back in the helm?

eh Cragg - see this as an example - Workunit details

THEY are called 'Wingmen' - You shall have to wait 'till THEY Finish their Tasks

(nothin' wrong w/ Your Results / Machines etc . . . ;)

---

BOINC Wiki . . .

Science Status Page . . .

THEY are called 'Wingmen' - You shall have to wait 'till THEY Finish their Tasks

(nothin' wrong w/ Your Results / Machines etc . . . ;)

[/quote]


Thank you for your quick reply, and the reassurance. I was not aware that results were dependent on others finishing the WU (until now, I haven't read the fine print in the Message Boards).

I guess that with everyone running Astropulse, and the "pause" it injects into the flow, that depending on when their machine "gets around" to crunching the shorter WUs (ie: creating a regular flow), that the delays can be quite long, especially if they pause one to run something in a different project. Evident by seeing the same stats in my signature over different days.

It's just painfull to see the little red arrows in the stats, instead of the little green ones.

Thanks again.

---

Thank you for your quick reply, and the reassurance. I was not aware that results were dependent on others finishing the WU (until now, I haven't read the fine print in the Message Boards).

I guess that with everyone running Astropulse, and the "pause" it injects into the flow, that depending on when their machine "gets around" to crunching the shorter WUs (ie: creating a regular flow), that the delays can be quite long, especially if they pause one to run something in a different project. Evident by seeing the same stats in my signature over different days.

It's just painfull to see the little red arrows in the stats, instead of the little green ones.

Thanks again.

Errm - unfortunately, yours stats will always have the same value. You're using a URL which always link to your latest figues. Have a look at your signature now and check back tomorrow - the same message will have the new figures (now that it appears the XML stats are being sent again). They look prety good as well (well, better than mine, anyway ;)).

BTW - I've had little blue ones for the past week!

---

Boincstats started updating yesterday FYI

---

Boincstats started updating yesterday FYI

Manually, onnce. crontab is still down.

73 Iztok

Boincstats started updating yesterday FYI

Manually, onnce. crontab is still down.

Why? Is there any problem?

---