[Cado-nfs-discuss] Cado-nfs-discuss Digest, Vol 73, Issue 5
Emmanuel.Thome at inria.fr
Tue Jul 10 11:18:43 CEST 2018
On Tue, Jul 10, 2018 at 04:01:22PM +0700, canny georgina wrote:
> when i tried to somewhat compute what time it took from (for example 1.3%
> to 1.4%), im confused since what the ETA shows is weird (for me).
> when I collect them all just like this:
> > ok (*1.3% => ETA Tue Jul 17 01:40:07 2018*)
> > ok *(1.3% => ETA Tue Jul 17 07:43:21 2018)*
> > ok (*1.4% => ETA Tue Jul 17 01:58:11 2018*)
> > ok (*1.4% => ETA Tue Jul 17 04:24:17 2018)*
> i'm confused about how to compute it when it's *01:40:07 ** going forward
> to this hour **( 07:43:21) and backward at this hour ( 01:58:11) and going
> forward again at this hour ( 04:24:17)*
No bug here.
The ETA is computed as
start_time + work_to_do * (now - start_time) / work_done.
At the beginning of the computation, you have enough irregularity in the
speed per workunit to account for jitter that adds up to hours or days
when multiplied by "work_to_do/work_done".
More information about the Cado-nfs-discuss