iraf.net

Hi Folks,What is the fastest and least error-prone method for copying FITS
images from disk to exabyte tape? I am composing instructions for
Keck observers to write FITS tapes, and note that there are at least
three ways in IRAF/STSDAS to accomplish the task:1) "fitscopy" appears simplest, but the help page warns that
"transfer to and from Exabyte tapes are considerably slow". 2) "stwfits" is more complicated than needed for our needs, being
specialized for HST data3) "wfits" would be fine, but I am wondering: a) what settings for bscale/bzero/scale/autoscale are
appropriate for copying a FITS file from tape to disk such
that the two files are identical? i.e., does "wfits scale-
autoscale-" produce a duplicate of the tape file? b) is this routine any more or less efficient at writing
Exabyte files than "fitscopy" c) for exabyte, is a blocking factor of 1 preferred over the
default blocking factor of 10? i.e., what are the implications
for I/O efficiency?Many thanks for your help! Aloha, Greg--
Gregory D. Wirth Email: wirth@keck.hawaii.edu
W. M. Keck Observatory
65-1120 Mamalahoa Hwy Phone: (808) 881-3866
Kamuela HI 96743 USA Fax : (808) 881-3897

Hello Greg;
I did some measurement using fitscopy and wfits with local
disk on a SUN sparc 10 running SUNOS 4.1.

> What is the fastest and least error-prone method for copying FITS
> images from disk to exabyte tape? I am composing instructions for
> Keck observers to write FITS tapes, and note that there are at least
> three ways in IRAF/STSDAS to accomplish the task:
>
> 1) "fitscopy" appears simplest, but the help page warns that
> "transfer to and from Exabyte tapes are considerably slow". Running many files with block=10 to an exabyte tape I got a
rate of around 70Kbytes/sec. One big file to the same tape a got 138Kbytes/sec. Tape i/o is the slow factor in here and I think using block=10
with help in this area.
>
> 2) "stwfits" is more complicated than needed for our needs, being
> specialized for HST data
>
> 3) "wfits" would be fine, but I am wondering:
>
> a) what settings for bscale/bzero/scale/autoscale are
> appropriate for copying a FITS file from tape to disk such
> that the two files are identical? i.e., does "wfits scale-
> autoscale-" produce a duplicate of the tape file?
>
In here you might want to try some tests. If you have a
file on disk that has scaled pixel values (i.e. BSCALE/BZERO not
the default values) and then you use wfits with scale-, auto-
you'll probably end up with a truncated pixel value data portion
on tape.> b) is this routine any more or less efficient at writing
> Exabyte files than "fitscopy" I think the efficiency is mostly tied to the tape drive.
>
> c) for exabyte, is a blocking factor of 1 preferred over the
> default blocking factor of 10? i.e., what are the implications
> for I/O efficiency? The newer exabyte model are more efficients in i/o and using a
bigger buffer size with the block=10 would help even more.
In summary, my advice is to use 'fitscopy' since it will not touch the
content of the header not the data portion of the FITS file. Using
wfits will modified the header by adding a modifying header cards
on tape.
>
> Many thanks for your help!
>
> Aloha,
>
> Greg
>
> --
> Gregory D. Wirth Email: wirth@keck.hawaii.edu
> W. M. Keck Observatory
> 65-1120 Mamalahoa Hwy Phone: (808) 881-3866
> Kamuela HI 96743 USA Fax : (808) 881-3897
> If you have more questions or comments about this, please let me know.Cheers
Nelson Zarate
Noao