170MB?

history
history
Joined: 22 Jan 05
Posts: 127
Credit: 7,573,923
RAC: 0
Topic 192296

O/A 2025 hours CST 10 January 2007, rig 75375 got a download from "Tthud"(our beloved server) of 11 each 15.7mb files and their associated grids. My disc went from 80mb to 250mb. I have been hauling water for this project for almost two years and cannot remember a prior instance of a download of this size. This seems to be a huge inefficiency to deliver WU's that crunch in under 30 minutes each. Anyone else out there a victim of "Wretched Excess"? Admins, I know you are busy stocking up on rubberbands and used raid arrays to keep Tthud wheezing along, but 170MB? These short WU's are killing us.

Regards-tweakster

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 1,932
Credit: 163,192,387
RAC: 166,954

170MB?

Quote:

O/A 2025 hours CST 10 January 2007, rig 75375 got a download from "Tthud"(our beloved server) of 11 each 15.7mb files and their associated grids. My disc went from 80mb to 250mb. I have been hauling water for this project for almost two years and cannot remember a prior instance of a download of this size. This seems to be a huge inefficiency to deliver WU's that crunch in under 30 minutes each. Anyone else out there a victim of "Wretched Excess"? Admins, I know you are busy stocking up on rubberbands and used raid arrays to keep Tthud wheezing along, but 170MB? These short WU's are killing us.

Regards-tweakster


I think it's more likely to be because we're getting close to the end of the S5R1 run. Normally, you just get one of these 15MB downloads, and then crunch lots of separate WUs from the same data.

But now, the bulk of the mainstream work will have been done, and much more of what we've got left will be going back to fill in the 'holes' caused by crunchers who miss deadlines or just drop out of the project - aggravated, of course, by the recent server dropouts and the 168-hour connection deferrals that can result. If you check the WUs you downloaded around the time of the big datafile downloads, I expect you'll find that a lot of them are re-issues after missed deadlines.

But you may have highlighted another factor in the recent server stress: probably too late to change the project design for S5R2, but it might be something to think about when planning for S6.

Edit - host 75375 doesn't seem to exist, and your computers are hidden so I can't tell whether it's a typing error or something. Do you want to re-post the host ID so the admins can check?

Fuzzy Duck
Fuzzy Duck
Joined: 3 Dec 05
Posts: 37
Credit: 930,924
RAC: 0

Richard, Do you help out

Message 59164 in response to message 59163

Richard,

Do you help out on all the BOINC message boards?

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 1,932
Credit: 163,192,387
RAC: 166,954

RE: Richard, Do you help

Message 59165 in response to message 59164

Quote:

Richard,

Do you help out on all the BOINC message boards?


LOL! No, mainly SETI and Einstein - but I've just started dabbling in SETI Beta and CPDN. You can tell from my post count that I'm not a heavy-duty volunteer, but I pitch in where I can (and when I need a displacement activity from the database application I'm supposed to be writing....)

history
history
Joined: 22 Jan 05
Posts: 127
Credit: 7,573,923
RAC: 0

Richard; Try rig 753745. My

Richard; Try rig 753745. My typing skills seem to have been impacted by frustration and lack of sleep. I appreciate your assistance. How about we reward you with a box of rubber bands and a collection of raid drivers on floppy discs all packaged up and sent via Fedex Ground (the premier "hide it in the bushes") delivery service?

Regards-tweakster

Richard Haselgrove
Richard Haselgrove
Joined: 10 Dec 05
Posts: 1,932
Credit: 163,192,387
RAC: 166,954

RE: Richard; Try rig

Message 59167 in response to message 59166

Quote:

Richard; Try rig 753745. My typing skills seem to have been impacted by frustration and lack of sleep. I appreciate your assistance. How about we reward you with a box of rubber bands and a collection of raid drivers on floppy discs all packaged up and sent via Fedex Ground (the premier "hide it in the bushes") delivery service?

Regards-tweakster


Yep, quite a lot of re-issues - mainly at the bottom half of the first page.

Another way to tell is to look at the 'Name' column in the BOINC task tabs - the first decimal number after the 'h1' or 'l1'. Every time it changes, you've got a new datapack download (unless you've done work in that sequence before, and you've already got the pack). Conversly, look how many you've got from 0381.0: that's all from one download, which doesn't look so bad!

[Don't worry about the rubber bands - our UK postal service leaves free offerings scattered all along the access path as they deliver the mail!]

Sou'westerly
Sou'westerly
Joined: 9 Jun 06
Posts: 57
Credit: 715,838
RAC: 0

RE: But you may have

Message 59168 in response to message 59163

Quote:
But you may have highlighted another factor in the recent server stress: probably too late to change the project design for S5R2, but it might be something to think about when planning for S6.


The easy way would be just to raise the criteria for what is a slow machine and qualifies for short work units. At the beginning of this run at a lot of people were complaining that they kept getting long WUs which were taking them too long to crunch. Better to be left with long WUs at the end of a run as that will cut the number of data files needing to be sent out by 10. dAVE.
Hindsight is a great thing!

KAMasud
KAMasud
Joined: 6 Oct 06
Posts: 13
Credit: 9,559,237
RAC: 2

:-) Yes, but these 15Mb files


:-) Yes, but these 15Mb files are a killer on dial up? one file takes at least an hour:-(
Regards
Masud.

Alinator
Alinator
Joined: 8 May 05
Posts: 927
Credit: 9,352,143
RAC: 0

Yep, EAH is tough enough for

Yep, EAH is tough enough for DU user's during the mainstream part of the run when they need new datapacks. I don't blame you for being annoyed a little during this final "clean up" phase, when you spend an hour dl'ing a datapack to crunch one or two results and then need to get another one. ;-)

But what can you do about it, other than NNW until S5R2? We've got to gid rid them in any event. :-)

Alinator

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.