The Question is:
in a shared scsi cluster configuration of 2 ds20 connected with gigabit
ethernetcards we see a network performance which is far below of what we
for example :
If a 6 MByte file is copied via ftp from one machine to the other the
transferrate is about 1,7 Mbyte/second. From performances seen with other
architectures ( NT; Netware; Unix ) I expected at least a 10 times higher
rate on a net without traffic.
The Disk subsystem cannot be the bottlenek, a restore from tape runs with
roughly 13Mbyte/second sustained.
Where can we start ?
The Answer is :
Try DECnet or SCS-level operations, to determine if this is FTP and IP,
or the underlying network connection. Ensure that process quotas are
set appropriately. Check for system parameter limitations, as well as
for sufficient free memory, as well as for spindle contention, as well
as for sufficient free pool... Find out what limit you are measuring,
in other words.