[comp.unix.admin] Exabyte.

bob@tekgen.BV.TEK.COM (Bob Wakehouse) (09/11/90)

According to documentation with my Tektronix XD88 Exabyte option, the tape
drive writes only full 8k tracks, in 1k blocks.  If less than 8k is ready
for writing, the remainder is padded to make an 8k track.  When stopping,
Exabyte writes an additional full 8k pad track.  So if the buffer doesn't
have 8k, then 1-7k tape is wasted on padding.  If data is then still not
ready, another 8k tape is wasted on padding, and the drive stops to await
more data.  In other words, each time the buffer does not have 8k ready to
write, 1-15k of tape is wasted on padding.

My docs claim the Exabyte buffer is 256k, or 4 times the common 1/4 inch
drive buffer.  Exabyte speed is listed as 3 times the 1/4 inch drive, so
it sounds reasonable that the docs say blocking factors of 10 or 20 are
at least as effective as very large blocking factors, for most operations.

The docs also claim that the tape drive will return EOT after completing
a write when the tape is about 90% full.  Subsequent write attempts return
ENXIO.

I quote, "Count on about 2.0 Gbytes from a 2.3 Gbyte tape".

Trying to put 31 files on tape, from across some roundabouty network 
connections, I ran into EOT problems at 1.5 gig of actual data.  I don't
know if the loss was due to excessive padding or just not-so-good tape
parameters to "dump" (XD88 "dump" is a bit different than standard "dumps",
as command options go).


Bob Wakehouse
bob@amati.tv.tek.com
Tektronix, Inc.
Beaverton, Oregon