Make sure you're either fully configured for IPV6 or disable it. I've had lots of performance problems with newer machines and IBMi versions. 6.1 and 7.1 have IPV6 enabled by default. If you aren't fully configured for IPV6 (ie an IPV6 DNS server on the machine) the IP stack spends an inordinate amount of time trying to find the other machine. It seems that time is spent on EVERY packet. Most of my experience is with Java applications which always favor IPV6 unless explicitly configured to prefer IPV4. I don't have much experience with the FTP client and server on IBMi, so this may not actually be a factor. But it's something to check.
From: midrange-l-bounces@xxxxxxxxxxxx [mailto:midrange-l-bounces@xxxxxxxxxxxx] On Behalf Of rob@xxxxxxxxx
Sent: Thursday, January 10, 2013 7:44 AM
Subject: Gb Ethernet
Shouldn't I be getting a lot better throughput?
Help me out with my math. I've got an Ethernet line on 1 lpar talking to an Ethernet line on another lpar. They both say:
Current line speed . . . . . . . . : 1G
Current duplex . . . . . . . . . . : *FULL
They are both on our same 10.17.6 subnet.
I FTP'd a sizeable file and got these results:
Size, in bytes, of save file: 13,458,505,728
Seconds to perform transmission: 2,222
iNav's Management central says lan utilization was minimal.
iNav's says percent busy of disk was minimal. (currently 2-7%) Source system has 64 disk arms.
Target system is a guest on the source. It has 6 equal "arms".
Shouldn't I be getting a lot better throughput? After all, 0.05 is not 1Gb.
I am not interested in any virtual ethernet backplane type solution due to some H/A concerns.