TIP: Click on subject to list as thread! ANSI
echo: novell
to: ALL
from: JEFF DUNLOP
date: 1996-06-30 19:11:00
subject: Heavy LAN traffic

I had a LAN fall flat on its face a couple times this week because of very
heavy load. There are about 25 PCs, ranging from 486/66 with Intel 
EtherExpress 16 to Pentium/75 with Intel EtherExpress Pro/10 PCI running a 
Foxpro 2.6 for Windows app off the server. There are three HP 4 Plus 
printers, 1 on the server, 2 on JetDirect cards operating as queue servers.
We've running into occasional problems when running reports from several PCs 
at the same time. It appears that the lan segment collapsed under the load 
because the Netware 3.12 server reports about 800,000 collisions and other 
assorted traffic-related errors.
I know that I can run multiple segments from the server, and that bandwidth 
can be improved in other areas. Indeed, we have a Compaq Proliant 300 150mhz 
server on order with an extra PCI NIC and all the performance bells and 
whistles. This will probably address our immediate needs, but we're still 
left with the probablistic nature of Ethernet.
Is there anything that can be done to make the LAN behavior more 
deterministic? For instance, any Netx settings that can cause the clients to 
be less aggressive on the segment if excessive collisions are detected? How 
much of an impact does the JetDirect have when printing graphical jobs? I'm 
thinking back to other LANs of similar size that I've seen statistics on, and 
some of them went for months or longer with fewer than a dozen collisions. I 
have a 28-user LAN on a WAN running on a single segment and an almost 
identical server that has never had a hiccup.
When I have a Foxpro developer telling me he needs twice the throughput for 
his one penny-ante application, I can't help but wonder if we should just 
make this application smarter. Any possibility that FoxPro's runtime thrashes 
when it gets denied a lock? Please don't suggest going C/S -- he can't meet a 
deadline with his ridiculous little language as it stands.
Finally, a faster server with additional IO channels will help throughput 
through the server, but it doesn't seem to me that would reduce the demands 
on the segment. Will a faster server aggravate our collision problems?
Jeff
--- GoldED/2 2.42.G0615
---------------
* Origin: DB/Soft Online - Sacramento, CA (916)927-2349 (1:203/16)

SOURCE: echomail via exec-pc

Email questions or comments to sysop@ipingthereforeiam.com
All parts of this website painstakingly hand-crafted in the U.S.A.!
IPTIA BBS/MUD/Terminal/Game Server List, © 2025 IPTIA Consulting™.