[Debburn-devel] [PATCH] Better ETA algorithm

[Nicolas George]

Hi.

To compute the estimated time, genisoimage uses only the start and current
time. When the output is piped directly to the burning process, there is
usually a significant time at the start before the burning starts; for a
BD-R, it can be more than five minutes. This cause the ETA to be completely
off in the first few minutes, and still widely inaccurate until more than
half the process, even if the progress is perfectly regular.

For the same reason, if a disk slows down near the end, possibly due to a
defect, the ETA will reflect it only very slowly.

Furthermore, when generating multisession images, the part that already
written on the medium is taken into account for the ETA, even though it did
not any time, and again, the ETA is thus widely inaccurate.

The attached patch uses an exponential moving mean on about 30 seconds and a
linear regression to compute the ETA. The graph at
http://www.normalesup.org/~george/tmp/genisoimage_eta.png
shows it fixes the first problem (simulated 12Go BD-R at speed 1; x is the
progress percent, y is the ETA in seconds since the start; red is the
unpatched version, green is the patched version), and hopefully it fixes all
problems the same way.

Regards,

-- 
  Nicolas George
-------------- next part --------------
A non-text attachment was scrubbed...
Name: 0001-genisoimage-change-the-ETA-algorithm.patch
Type: text/x-diff
Size: 4016 bytes
Desc: not available
URL: <http://lists.alioth.debian.org/pipermail/debburn-devel/attachments/20110608/e3ba0310/attachment.patch>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: Digital signature
URL: <http://lists.alioth.debian.org/pipermail/debburn-devel/attachments/20110608/e3ba0310/attachment.pgp>