Dino wrote:
Does anyone have any ballpark numbers for actual throughput performance of
FTP using Hipersockets
Running either single or multiple FTPs from LPAR #1 to LPAR #2 we are only
seeing around 25MB/sec in total.
The DASD involved is all FICON connected DMX-3 which is easily capable of
transferring faster than this.
There is also ample CPU capacity available.
Any thoughts on what may be constraining throughput would greatly
appreciated.
Thanks, Dino
We just recently setup hipersockets for communication test between z/OS
and Linux (under VM) and I was so disappointed when I could not get any
more than 25MB/sec or so.
I spent a week beating my head against the wall trying to figure out why
I was getting so "slow" through put on ftp between z/OS and Linux
using Hipersockets. I tweaked MTU, started at 56K went down to 32K,
then do 16K and no difference (put it back to 56K and left it). Then I
started playing with buffer sizes on both the z/OS and Linux side so I
could to try and get bigger Window sizes.
IIRC the best I could ever get was 40MB/sec and on average it was
25MB/sec. Still very disapointing as this was equal to untuned 1Gbps
Ethernet.
I finaly found in a red book that FTP, even using hipersockets is not
going to get much better than 40 MB/sec. The issue apparently a fixed
buffer size within FTP.
I did one test where I ran 3 FTP sessions at once and each one got
30-35MB/sec. So, it did seem that FTP was the limiting factor. I never
tested with more that 3 sessions, because the DASD setup on the Linux
side DASD I/O was going to be the limiting factor. But I do know that
with more DASD (actually just spread across more volumes) I could have
done more sessions all doing 30-35MB/sec.
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to [EMAIL PROTECTED] with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
