Index  | Recent Threads  | Unanswered Threads  | Who's Active  | Guidelines  | Search
 

Quick Go »
No member browsing this thread
Thread Status: Active
Total posts in this thread: 121
Posts: 121   Pages: 13   [ Previous Page | 4 5 6 7 8 9 10 11 12 13 | Next Page ]
[ Jump to Last Post ]
Post new Thread
Author
Previous Thread This topic has been viewed 99285 times and has 120 replies Next Thread
KWSN - A Shrubbery
Master Cruncher
Joined: Jan 8, 2006
Post Count: 1585
Status: Offline
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

Roundup, if you read the thread, [...].

Yes Sir. I request a hard but fair punishment wink


In the true spirt of the Knights Who Say Ni!, we will subject you to the harshest punishment of them all....


The comfy chair!!!
----------------------------------------

Distributed computing volunteer since September 27, 2000
[Mar 10, 2010 6:18:44 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Ingleside
Veteran Cruncher
Norway
Joined: Nov 19, 2005
Post Count: 974
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

I've posted a provisional update of Matrix in the Start Here forum and added a few columns which I'm hoping to fixed up with permanent data.

Standard/Repair Deadlines

A Type 20/8
B Type 8/3.2
C Type 10/4

Shouldn't it be 8/8 for A, and 7/2.8 for B, according to Kevin Reeds post.

Reliability Return period and Valid Result Series columns were added. Previously knew of 4 days and 15 valid sequential results, it not 100% transparent to me what the DDDT-2 A/B are, thinking C will be same as all other WCG projects.

There's not separate "reliable"-parameters for each sub-project, meaning a computer will either be reliable in all WCG-projects, or in none of them. Also, it doesn't matter in which sub-projects you'll get the valid tasks (or invalids), so you can example have 10 HCMD2, 15 HFCC, 20 Rice, 8 HCC, 12 HPF2, 9 FAH and 3 beta's to get the required 77 valid tasks to become "reliable".


So, a new computers needs 77 valid tasks in a row to drop below the max_error_rate of 0.002. If has a single validation-error, the computer needs atleast 77 valid tasks in a row afterwards to get below 0.002 again...

On top of this you'll need the max 2 days average turnaround-time, but even if it is higher, it's much easier to drop below 2 again than to crunch the 77 needed to decrease the error-rate.
----------------------------------------


"I make so many mistakes. But then just think of all the mistakes I don't make, although I might."
----------------------------------------
[Edit 1 times, last edit by Ingleside at Mar 10, 2010 7:25:27 PM]
[Mar 10, 2010 7:23:42 PM]   Link   Report threatening or abusive post: please login first  Go to top 
roundup
Veteran Cruncher
Switzerland
Joined: Jul 25, 2006
Post Count: 824
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

[ot]

In the true spirt of the Knights Who Say Ni!, we will subject you to the harshest punishment of them all....

The comfy chair!!!

Nooooo please, you can do anything but please please not the comfy chair crying beat up
[/ot]
[Mar 10, 2010 7:27:16 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Sekerob
Ace Cruncher
Joined: Jul 24, 2005
Post Count: 20043
Status: Offline
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

Suggest you wait for the final matrix. 20/8 means as it says in the header, Standard Deadline A type 20 days, Repair Deadline 8 days. FURTHER, IF you'd read, you'd seen a note saying the 7 was increased to 8, consequently 2.8 would then become 3.2. The rest will resolve as an when the numbers have confirmed.

Plz, no injections how you may have gotten it wrong or I got it wrong... there's enough to read already.

edit: post quote v.v. B-Type
Ingleside, the avg turnaround for the host is 2 days...but the work units will be given 8 (updated originally it was 7) days to return the work. Let me know if that is what you are curious about?

-Uplinger
[Mar 8, 2010 8:17:00 PM]


edit: what better value to use than the one on a received B-Type download: 7 Days it is.
http://www.worldcommunitygrid.org/forums/wcg/...28656_lastpage,yes#271142
----------------------------------------
WCG Global & Research > Make Proposal Help: Start Here!
Please help to make the Forums an enjoyable experience for All!
----------------------------------------
[Edit 2 times, last edit by Sekerob at Mar 11, 2010 7:56:46 AM]
[Mar 10, 2010 7:34:41 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

Now you've got me curious about how reliable my machine is rated.

Is there anywhere you can find out your reliability rating and average turn-around times? I had a look around but couldn't find it.

Also, how long a period is the average turn-around time calculated over? Forever? the last n months/work-units?
[Mar 15, 2010 12:47:12 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Ingleside
Veteran Cruncher
Norway
Joined: Nov 19, 2005
Post Count: 974
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

Now you've got me curious about how reliable my machine is rated.

Is there anywhere you can find out your reliability rating and average turn-around times? I had a look around but couldn't find it.

In BOINC-projects, the average turnaround-time is displayed on the show_host.php-page for each computer, but WCG doesn't include this page so you can't see it here.

The reliability-rating isn't shown anywhere...

A computer is "reliable", if it gets assigned tasks with decreased deadline...

Also, how long a period is the average turn-around time calculated over? Forever? the last n months/work-units?

if you've got a task timing-out, the average will be set to how many days the original deadline on this task was. So, even if no-one gets more than 8 days deadline on an A-task, if you miss-out on the 8-day deadline on the A-task your average will be set to 20 days.

As long as you don't exceed the deadline on any tasks, you can maybe say it's "forever", but the averaging is a 70/30-weight on old/new, so even if you're sitting with a 10-day average it will drop below 2 days after 7 tasks taking 1 day.
----------------------------------------


"I make so many mistakes. But then just think of all the mistakes I don't make, although I might."
----------------------------------------
[Edit 1 times, last edit by Ingleside at Mar 15, 2010 5:50:58 PM]
[Mar 15, 2010 5:51:25 AM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

I have some machines that are not always cruching WCG, but when they are they get short work so they must be rated reliable. If they then go offline for a week as far as WCG is concerned do they keep their rating when they next contact WCG for a wu? They wouldn't have returned work late just been out of touch with the grid for a week not requesting any work units.
[Mar 15, 2010 2:02:29 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Former Member
Cruncher
Joined: May 22, 2018
Post Count: 0
Status: Offline
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

Hello w0en,
As far as I know, your computers will still be considered reliable. The test is how a computer handles work assigned to it, not how often it requests work.

Lawrence
[Mar 15, 2010 2:18:46 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Ingleside
Veteran Cruncher
Norway
Joined: Nov 19, 2005
Post Count: 974
Status: Offline
Project Badges:
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

Hello w0en,
As far as I know, your computers will still be considered reliable. The test is how a computer handles work assigned to it, not how often it requests work.

Lawrence

Correct. smile

Just adding a little more info...

Except if a task times-out and isn't returned by the deadline, all the updating of average turnaround-time is done by the Validator, so not crunching WCG for a month or something won't have any effects on the turnaround-time here at WCG.

Since average turnaround-time is adjusted by the Validator, there can be some delay waiting on your wingman before it's updated. wink

For new computers the average turnaround-time starts at zero, so you'll only need 1 validated task to get average turnaround-time within limits. But you'll also need to get low enough host.error_rate to be rated reliable, and for this you'll need atleast 77 validated tasks...


BTW, for the users that does somehow manage getting an A-task, but uses more than 48 hours to crunch it, this normally won't lead to unreliable computer. A computer with 1-day average can return one 100-hour-task and still be reliable, or two 71-hour-tasks. Also, even if does return more A-tasks, chances are they won't be validated at once, so you'll get some other work validated between the A-tasks. And, as long as the "unreliable"-rating is only due to turnaround-time, it takes very few fast turnaround-tasks to get the average below 2 again. cool
----------------------------------------


"I make so many mistakes. But then just think of all the mistakes I don't make, although I might."
[Mar 15, 2010 5:48:27 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Sekerob
Ace Cruncher
Joined: Jul 24, 2005
Post Count: 20043
Status: Offline
Reply to this Post  Reply with Quote 
Re: DDDT-2 Work Unit distribution updated

If has a single validation-error, the computer needs atleast 77 valid tasks in a row afterwards to get below 0.002 again...


Did you have a lapse, or where in the standard BOINC world is that documented to apply to WCG too? Remember, this is WCG where old world and new world sublimate into something overall vastly superior. Retention rate of 34% of BOINC members versus 16.9% for all the rest of BOINC standard projects is in that my reference, that's more than twice without a Reliability rating page.
----------------------------------------
WCG Global & Research > Make Proposal Help: Start Here!
Please help to make the Forums an enjoyable experience for All!
[Mar 15, 2010 6:41:38 PM]   Link   Report threatening or abusive post: please login first  Go to top 
Posts: 121   Pages: 13   [ Previous Page | 4 5 6 7 8 9 10 11 12 13 | Next Page ]
[ Jump to Last Post ]
Post new Thread