Abysmal IPX Routing performance
Ross Vandegrift
My boss recently got a cable modem installed to replace his ISDN
connection to the office. I've been configuring VPN stuff so he'll be
able to login to our Netware 3.12 network from his house via standard
Netware DOS clients (yes, we use DOS).
Currently, the setup looks like this:
Primary office network (at office) ->
GNU/Linux router running a PPP VPN and ipxripd (at office) ->
GNU/Linux router running PPP VPN/ipxripd (at boss's house) ->
IPX Network at boss's house
This setup works, but is awfully slow. The internet connection at the
office is a 512k frac-T1. The cable modem has ~1M downstream. TCP/IP
transfers over the VPN easily fill our T1. IPX transfers run at about
7k/sec.
We've tried using Novell's IPTUNNEL.NLM at both ends, but had the same
results. All of the machines involved are more than capable of filling
our 512k T1.
Is there some grevious mis-design here? It seems to make sense to me,
but I'm 99% Unix geek.
Would we see a performance gain in building two modern (v5 or 6) Netware
machines to speak IPX to Ethernet clients and IP over the VPN?
Thanks in advance for and insight or advice,
Ross Vandegrift
ro...@willow.seitz.com
Dave Kearns-NSCV
news:3D2DDD4...@willow.seitz.com...
>
> Would we see a performance gain in building two modern
> (v5 or 6) Netware machines to speak IPX to Ethernet clients
> and IP over the VPN?
>
Yes. But you could also go ahead and eliminate ipx entirely after the
upgrade. IPX generally creates multiple small packets each of which
must be ACKed - its slow over a dial-up WAN.
--
Dave Kearns
Novell Support Connection Volunteer