v4 LL result assignment times graph

Old man PrimeNet · 2006-02-16, 18:55

This graph of 19538 recent LL results shows 98% are received within 32 weeks (224 days) of assignment, with the peak assignment age at 6 weeks. Factor results have a smaller, similar curve peaking at 1 week.

v5 PrimeNet will manage the long, thin tail of the LL result time curve by more judiciously assigning work according to computer capabilities. The tail will still be long, but thinner.

Click image for larger version

Name: v4_LL_result_times_weeks.jpg
Views: 139
Size: 28.9 KB
ID: 1040

garo · 2006-02-16, 19:12

This is fascinating information that I've always been interested in knowing! It would also be interesting to know what proportion of expired exponents return a result and when. This should help us fix the "expiry" date more accurately in v5.

BTW, is this graph for first-time test only or both first-time and second time. it would be useful to break this into three graphs for first-time, DC and 33Ms. And also perhaps according to exponent size.

Old man PrimeNet · 2006-02-16, 23:02

Correction: the main results arrival peak is centered on week 4, not 6 as stated above.

Below is a component breakdown chart, though for only 7161 LL results received in the last 24 hours that were assigned up to a year ago.

Right, garo, this information does in fact inform our v5 assignment expiration policy planning. But for that purpose the graphs do not need to differentiate expired assignments, which are included.

Assignment time is mostly a function of exponent size, computer speed/model and availability, and we can calculate how long a LL result should take for a given computer fairly accurately when we have good computer data. The only reason we expire assignments is to get results back faster. The v5 system will have the data to manage different expiration times for each computer based upon their characteristics, performance history and reliability. Other v5 possibilities include account users and teams configuring their own expiration policies, within reasonable limits.

2006-02-16, 18:55	#1
Old man PrimeNet Jan 2003 Altitude>12,500 MSL 101₁₀ Posts	v4 LL result assignment times graph This graph of 19538 recent LL results shows 98% are received within 32 weeks (224 days) of assignment, with the peak assignment age at 6 weeks. Factor results have a smaller, similar curve peaking at 1 week. v5 PrimeNet will manage the long, thin tail of the LL result time curve by more judiciously assigning work according to computer capabilities. The tail will still be long, but thinner.

2006-02-16, 19:12	#2
garo Aug 2002 Termonfeckin, IE 3·919 Posts	This is fascinating information that I've always been interested in knowing! It would also be interesting to know what proportion of expired exponents return a result and when. This should help us fix the "expiry" date more accurately in v5. BTW, is this graph for first-time test only or both first-time and second time. it would be useful to break this into three graphs for first-time, DC and 33Ms. And also perhaps according to exponent size. Last fiddled with by garo on 2006-02-16 at 19:14

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
The Euclid-Mullin graph	sean	Factoring	59	2016-05-29 08:26
Manual submission of automatic assignment result	tichy	PrimeNet	4	2010-12-17 09:57
Progress Report - Graph?	vaughan	No Prime Left Behind	9	2009-07-17 07:32
Simple graph to predict Primes	kaldor	Homework Help	11	2007-04-12 13:53
Graph of server assignments	tha	PrimeNet	1	2006-12-14 00:24

2006-02-16, 23:02	#3
Old man PrimeNet Jan 2003 Altitude>12,500 MSL 101 Posts	Correction: the main results arrival peak is centered on week 4, not 6 as stated above. Below is a component breakdown chart, though for only 7161 LL results received in the last 24 hours that were assigned up to a year ago. Right, garo, this information does in fact inform our v5 assignment expiration policy planning. But for that purpose the graphs do not need to differentiate expired assignments, which are included. Assignment time is mostly a function of exponent size, computer speed/model and availability, and we can calculate how long a LL result should take for a given computer fairly accurately when we have good computer data. The only reason we expire assignments is to get results back faster. The v5 system will have the data to manage different expiration times for each computer based upon their characteristics, performance history and reliability. Other v5 possibilities include account users and teams configuring their own expiration policies, within reasonable limits.