Why?

Message boards : Number crunching : Why?
Message board moderation

To post messages, you must log in.

AuthorMessage
Profile AllenIN
Volunteer tester
Avatar

Send message
Joined: 5 Dec 00
Posts: 292
Credit: 58,297,005
RAC: 311
United States
Message 646872 - Posted: 23 Sep 2007, 1:19:29 UTC

Hello:

I would like to know why I am crunching a WU that is due in December, instead of one that is due in a week. Does anyone know?

Allen
ID: 646872 · Report as offensive
archae86

Send message
Joined: 31 Aug 99
Posts: 909
Credit: 1,582,816
RAC: 0
United States
Message 646877 - Posted: 23 Sep 2007, 1:29:47 UTC - in response to Message 646872.  


I would like to know why I am crunching a WU that is due in December, instead of one that is due in a week. Does anyone know?
luck of the draw at the moment you downloaded.

Ones expected to take a lot of computing time get a deadline long after your download, ones expected to take very little effort get a short deadline.

They are assigned a deadline, which is personal to you, at the moment they are assigned to you for download.

ID: 646877 · Report as offensive
Profile Uli
Volunteer tester
Avatar

Send message
Joined: 6 Feb 00
Posts: 10923
Credit: 5,996,015
RAC: 1
Germany
Message 646884 - Posted: 23 Sep 2007, 1:42:28 UTC

Boinc will adjust itself automaticly and should put this WU on hold so to speak. If you are worried about not being able to finish the newer WUs ontime, you could suspend it. Should you decide to go this route, you will not get any more work, while you have suspends.
I don't know how to add links yet to other threads regarding these issues, so maybe someone can help him with providing this.
Pluto will always be a planet to me.

Seti Ambassador
Not to late to order an Anni Shirt
ID: 646884 · Report as offensive
Profile AllenIN
Volunteer tester
Avatar

Send message
Joined: 5 Dec 00
Posts: 292
Credit: 58,297,005
RAC: 311
United States
Message 646885 - Posted: 23 Sep 2007, 1:44:17 UTC - in response to Message 646877.  


I would like to know why I am crunching a WU that is due in December, instead of one that is due in a week. Does anyone know?
luck of the draw at the moment you downloaded.

Ones expected to take a lot of computing time get a deadline long after your download, ones expected to take very little effort get a short deadline.

They are assigned a deadline, which is personal to you, at the moment they are assigned to you for download.


Interesting. The one that is due in December is about 6 hours to run and the one due next month is about 3 hours to run. I wouldn't think that the program would think it was necessary to run the December unit this soon over the 3 hour unit. Oh well, as long as it makes some sense to someone.

What's really odd is that I have 167 units in my queue and they will all be processed within the next couple of days, so it seems really odd that the picking and choosing even happens at all.

Guess I just don't understand the underpinnings of the software involved.

Allen

ID: 646885 · Report as offensive
Profile AllenIN
Volunteer tester
Avatar

Send message
Joined: 5 Dec 00
Posts: 292
Credit: 58,297,005
RAC: 311
United States
Message 646887 - Posted: 23 Sep 2007, 1:46:32 UTC - in response to Message 646884.  

Boinc will adjust itself automaticly and should put this WU on hold so to speak. If you are worried about not being able to finish the newer WUs ontime, you could suspend it. Should you decide to go this route, you will not get any more work, while you have suspends.
I don't know how to add links yet to other threads regarding these issues, so maybe someone can help him with providing this.



That's interesting. I didn't know that future d/ls of WU's halted as long as you had suspended WU's. Thanks for the info!

Allen

ID: 646887 · Report as offensive
Profile Uli
Volunteer tester
Avatar

Send message
Joined: 6 Feb 00
Posts: 10923
Credit: 5,996,015
RAC: 1
Germany
Message 646891 - Posted: 23 Sep 2007, 1:53:26 UTC

You are welcome Allen. I am trying to read thru the threads new and old to learn more on how things work.
Pluto will always be a planet to me.

Seti Ambassador
Not to late to order an Anni Shirt
ID: 646891 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 646943 - Posted: 23 Sep 2007, 3:25:08 UTC - in response to Message 646872.  

Hello:

I would like to know why I am crunching a WU that is due in December, instead of one that is due in a week. Does anyone know?

Allen

Unless there's a deadline problem, the BOINC core client crunches WUs in the order the tasks were assigned.
                                                                  Joe
ID: 646943 · Report as offensive
1mp0£173
Volunteer tester

Send message
Joined: 3 Apr 99
Posts: 8423
Credit: 356,897
RAC: 0
United States
Message 646947 - Posted: 23 Sep 2007, 3:30:14 UTC - in response to Message 646885.  


I would like to know why I am crunching a WU that is due in December, instead of one that is due in a week. Does anyone know?
luck of the draw at the moment you downloaded.

Ones expected to take a lot of computing time get a deadline long after your download, ones expected to take very little effort get a short deadline.

They are assigned a deadline, which is personal to you, at the moment they are assigned to you for download.


Interesting. The one that is due in December is about 6 hours to run and the one due next month is about 3 hours to run. I wouldn't think that the program would think it was necessary to run the December unit this soon over the 3 hour unit. Oh well, as long as it makes some sense to someone.

What's really odd is that I have 167 units in my queue and they will all be processed within the next couple of days, so it seems really odd that the picking and choosing even happens at all.

Guess I just don't understand the underpinnings of the software involved.

Allen

To expand on what Joe said.....

BOINC looks at the work you have in your queue, the likely amount of computing time available to BOINC, and the deadlines.

If there is no pressure, work is done in the order it was assigned to you.

If there is a chance that deadlines might be missed, BOINC does those work units that are "at risk" first.

Since as you say all 167 units in your queue will finish well ahead of deadlines, there is no need to do them in any particular order.
ID: 646947 · Report as offensive
Profile AllenIN
Volunteer tester
Avatar

Send message
Joined: 5 Dec 00
Posts: 292
Credit: 58,297,005
RAC: 311
United States
Message 646959 - Posted: 23 Sep 2007, 3:49:56 UTC - in response to Message 646943.  

Hello:

I would like to know why I am crunching a WU that is due in December, instead of one that is due in a week. Does anyone know?

Allen

Unless there's a deadline problem, the BOINC core client crunches WUs in the order the tasks were assigned.
                                                                  Joe



I guess I am curious then, why a date further out in the future would be assigned before a date closer to todays date? Now I guess I am wondering how the dates are assigned in the first place.

Why would a due date in December be assigned for a WU being issued in September and yet a WU assigned a week later in September might have an earlier due date?

Sometimes I wonder about the darnedest things........grin

Allen
ID: 646959 · Report as offensive
Profile AllenIN
Volunteer tester
Avatar

Send message
Joined: 5 Dec 00
Posts: 292
Credit: 58,297,005
RAC: 311
United States
Message 646960 - Posted: 23 Sep 2007, 3:52:43 UTC - in response to Message 646947.  


To expand on what Joe said.....

BOINC looks at the work you have in your queue, the likely amount of computing time available to BOINC, and the deadlines.

If there is no pressure, work is done in the order it was assigned to you.

If there is a chance that deadlines might be missed, BOINC does those work units that are "at risk" first.

Since as you say all 167 units in your queue will finish well ahead of deadlines, there is no need to do them in any particular order.


Okay, that seems reasonable, but why not assign earlier dates to the earlier issues and later dates to the later issued WUs in the first place? Seems sorta hit and miss to me on how the WU return dates are determined.

Allen

ID: 646960 · Report as offensive
Astro
Volunteer tester
Avatar

Send message
Joined: 16 Apr 02
Posts: 8026
Credit: 600,015
RAC: 0
Message 646964 - Posted: 23 Sep 2007, 4:06:17 UTC
Last modified: 23 Sep 2007, 4:08:15 UTC

All wus start life on a disc (formerly a tape), the splitter reads the recorded data in order, each 107 seconds is split into 256 sub bands, then copies of each are put into the database as "wus ready to send". The deadline comes from the AR (angle range) of each wu, and the angle range is dependent upon the speed of the gondola moving over the Arecibo dish, elevation, and rotational speed of the earth at that elevation. So, each file the splitters split can vary in Angle range. (I.E vary in deadlines), so one might be a quicky, the next a sloooooow one with longer deadline. Unless Boinc thinks you're in trouble of making a deadline, it'll proceed to process them in the order they were handed out.

NOTE: I'm sure they're more to the process, than I've listed, but was written to get the user to understand "the gist" of it.
ID: 646964 · Report as offensive
Josef W. Segur
Volunteer developer
Volunteer tester

Send message
Joined: 30 Oct 99
Posts: 4504
Credit: 1,414,761
RAC: 0
United States
Message 647109 - Posted: 23 Sep 2007, 13:27:42 UTC - in response to Message 646959.  

Hello:

I would like to know why I am crunching a WU that is due in December, instead of one that is due in a week. Does anyone know?

Allen

Unless there's a deadline problem, the BOINC core client crunches WUs in the order the tasks were assigned.
                                                                  Joe



I guess I am curious then, why a date further out in the future would be assigned before a date closer to todays date? Now I guess I am wondering how the dates are assigned in the first place.

Why would a due date in December be assigned for a WU being issued in September and yet a WU assigned a week later in September might have an earlier due date?

Sometimes I wonder about the darnedest things........grin

Allen

Crunch time is very dependent on how fast the telescope was moving, and deadlines are assigned based on how long a host with a Whetstone benchmark of 33.33 MIPS might take to do the work. This Classic FAQ entry shows what processing is done at different angle ranges. Very early in the development of setiathome_enhanced it became apparent that its more thorough analysis magnified the time differences. A set of time estimate formulas were developed which currently indicate that the longest crunching at 0.2259 angle range takes about 13 times as long as the shortest ones at 1.13 angle range and above. The shortest ones get a deadline which allows 8.68 days for slow crunchers, the longest ones allow nearly 113 days.
                                                               Joe
ID: 647109 · Report as offensive
Profile AllenIN
Volunteer tester
Avatar

Send message
Joined: 5 Dec 00
Posts: 292
Credit: 58,297,005
RAC: 311
United States
Message 647253 - Posted: 23 Sep 2007, 17:18:50 UTC - in response to Message 646964.  

All wus start life on a disc (formerly a tape), the splitter reads the recorded data in order, each 107 seconds is split into 256 sub bands, then copies of each are put into the database as "wus ready to send". The deadline comes from the AR (angle range) of each wu, and the angle range is dependent upon the speed of the gondola moving over the Arecibo dish, elevation, and rotational speed of the earth at that elevation. So, each file the splitters split can vary in Angle range. (I.E vary in deadlines), so one might be a quicky, the next a sloooooow one with longer deadline. Unless Boinc thinks you're in trouble of making a deadline, it'll proceed to process them in the order they were handed out.

NOTE: I'm sure they're more to the process, than I've listed, but was written to get the user to understand "the gist" of it.


Thanks Astro! I think I'm starting to get the gist of it now. Not all is so simple in OZ...grin

Allen
ID: 647253 · Report as offensive
Profile AllenIN
Volunteer tester
Avatar

Send message
Joined: 5 Dec 00
Posts: 292
Credit: 58,297,005
RAC: 311
United States
Message 647255 - Posted: 23 Sep 2007, 17:25:43 UTC - in response to Message 647109.  


Crunch time is very dependent on how fast the telescope was moving, and deadlines are assigned based on how long a host with a Whetstone benchmark of 33.33 MIPS might take to do the work. This Classic FAQ entry shows what processing is done at different angle ranges. Very early in the development of setiathome_enhanced it became apparent that its more thorough analysis magnified the time differences. A set of time estimate formulas were developed which currently indicate that the longest crunching at 0.2259 angle range takes about 13 times as long as the shortest ones at 1.13 angle range and above. The shortest ones get a deadline which allows 8.68 days for slow crunchers, the longest ones allow nearly 113 days. [pre] Joe


Thanks for even more info. As I mentioned to Astro, it's much more complicated than I had imagined. I'm still rather surprised that it chooses to do a WU due in December, before one due in a few days, since my processing time is so short on both of them, but I guess until you see the BIG picture, it will never really make good sense on the shallow level that we are privy to.

I thought that perhaps the software was able to send really long WU's to clients with really fast machines and the fast WU's to people with slower machines, in order to keep due dates acceptable to the program. Just shows I should quit thinking about things like this.......hahaha

Allen

ID: 647255 · Report as offensive

Message boards : Number crunching : Why?


 
©2024 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.