I'm not a super-network genius so please bare with
me.
We've got a foxpro application (client/server) that runs verrrry slow over
everyone's workstations (win98 workstations, win2k server). If I run
the same program from our one win2k client (used for terminal services), it's
fast. I ran ethereal to see if it was a network issue, and it seems
to be. A small transaction generates 3000 packets/frames on the win2k
box, but 8000 on the win98. And it gets worse the larger the
transaction (7 sec compared to 6 min. for larger trans.) The bulk of these
extra packets generated by the win98 conversation are 'flush
requests'. (the win2k box sends no such requests) (also the win98 box
does raw reads/writes, whereas the win2k uses AndX commands) Some further
digging revealed that the win2k box receives NT_STATUS error messages, and the
win98 box receives DOS_ERRORs. Is this a LanMan1.0 <-> NT LanMan 0.12
issue? (I'm no expert on this stuff, just what I've read about since
trying to diagnose the difference)
This is an old problem I sort of forgot about because it's not
a huge issue around here and I had no idea where to find an answer, but then I
recently remembered this list, so I thought I'd give it a try, and dug up the
old message I asked to a forum at the time.
Any help would be greatly appreciated, and I have packet
captures if needed.
thanx,
Curtis
|