rick@wiau.medical-biophysics.manchester.ac.uk (01/02/91)
We have had some discussion of this over her in the UK. Here is a summary
of what I have learned. I have not had a chance to get inside the software
or hardware and really look around.
126 IS NOT MAGIC!!!!
On my sys-admin course I was told increading the blocking factor speeds up
tape writes, and hay presto it does.
You can use any blocking factor they are interchangable.
1 block = 512 bytes (I am pretty sure but don't quote me)
The blocking factor say how many blocks (512bytes) to read/write at a time.
the maximum blocking factor you can use is
avaliable memory I can use
Blocking factor = --------------------------
block size * 2
which on a Sun-2 works out (apparently) at 126 (which is where 126 comes
from)
what happens when a 1/4" tape drive is used is:
bring tape upto speed
read/write blocking-factor blocks
slow down tape
rewind it a bit
etc.
so the bigger blocking factor the less overhead.
NOTES
56 is "magic" for exabytes otherwise you don't use all the track (one
exabyte track = 56 blocks)
I use 500 for 1/4" tape systems which is about twice the speed of 126
Our 1/2" drive the max blocking factor is 48 as it only has 24Kbytes of
internal memory (it is an M4 9903 drive)
As you increase you blocking factor it will use more system memory and
slow other thing down, so use a blocking factor worked out on the
avaliable memory you can use not total "on board" memory
As far as I can see 126 has got noting to do with how much data you can
get over a SCSI bus in one go
[[Ed's Note: I could be way out in wonderland with this - most of my tape
experience comes from the world of IBM and VM - but one thing that
different blocking factors meant in that day and age was how much data
physically got written on the tape in a block. The tape consisted of bands
of data segments separated by inter-data spaces. So, if you made your
blocking factor huge, then you could fit a lot more data on the tape,
*BUT* if one block of data was corrupted on your tape, you stood to loose
much more than if you wrote a bunch of tiny blocks. Again, I suspect that
this is probably different under SunOS, but I'd like to hear a real guru's
opinion. -bdg]]
Hope all this helps
RICK DIPPER, Wolfson Image Analysis Unit, rick@uk.ac.man.mb.wiau Department of Medical Bio-Physics, University of Manchester 061-275-5158