 |
 |
 |
 |
(1) HP340 w/40M 5.6 k/s
(2) HP340 w/200M 7.9 k/s
(3) HP340 w/200M FCS OFF 19.8 k/s
(4) MicroVAX II 40.2 k/s
(k/s = kilobytes per second). In (1) and (2), files were sent from an
HP340 to a MicroVAX II with FCS conversion ON. (3) is the same as (2),
except that FCS conversion is OFF. (4) is a VAX to VAX transfer. All
transfers were done with BD's FACSnet software (ftp). Rates were computed
based on the time it took to transfer 16 158k files, and as such, measure
the overall transfer through-put rate.
Some observations: (A) the 200 Megabyte hard drive from Bering speeds up
ftp transfer *through-put* by ~41% when FCS conversion is ON [(2) v. (1)],
(B) through-put is 2.5 times greater with FCS conversion OFF (3) v. ON (2),
and (C) even the fastest through-put rate (40.2 k/s) is far below the
theoretical limit (10 Mbits/s = 1280 k/s) for ethernet.
Some explanations: FACSnet makes a temporary file on the HP hard disk when
FCS conversion is ON. It is the temporary file that transfers across the
network. When FCS conversion is OFF, FACSnet doesn't waste its time making
temporary files. This explains (B). The 40M hard drives (HP9153C) are much
slower (in terms of disk access time, mean transfer rate, and other
technical specifications unfit for print here 8-{) than the 200M drives
made by Bering. Since FCS conversion is very disk-intensive (just listen to
your drive go clunk-clunking), the Bering speeds up through-put, thus
explaining (A). It is more difficult to account for (C). The net traffic
around here isn't *that* bad, and all machines are connected through the
same Local Network Interface (i.e. the machines are very 'close' to each
other).
/\/\/\_ Eric Van Buren vanburen%flovax.dnet@rocdec.roc.wayne.edu
\ \ \-' Immuno/Micro 313 577 1009 voice
\_^_/ Wayne State U. 313 577 1155 fax
|
|
|
|