We are having a number of issues with our Internet connection which
may be
related to our setup which might have a few minor errors but nothing
major
or may be related to the way in which the connection is provided. I
know
this is a long post but please bear with me, the relevance to BM is
further down.

Our setup is as follows - 2 subnets 172.16.46.0/24 and 172.16.20.0/22

separated by a NW50SP6A server running DHCP for both subnets and
routing
between the 2 (no static routes defined). All machines on both subnets

connect to a NW6SP2 with BM3.7 running DNS, HTTP proxy and NAT on the

public interface. Private IP 172.16.20.1, Public IP 172.16.44.1.

Global DHCP Options are
DNS: 172.16.20.1
IP Forwarding enable/disable: True
Perform router discovery: True
WWW server: 172.16.20.1
NDS Tree Name: The correct tree name

For 172.16.20.x DHCP Option
Router: 172.16.20.1

For 172.16.46.x DHCP Option
Router 172.16.46.2 (Which is the router)

The public side of the BM server then connects to a CISCO switch which

connects via fibre to a switch across town which goes through a CISCO

proxy and firewall. I have no further details of the hardware or setup
on
the public side as this was all installed as part of a broader network

connecting some 15 organisations at current although I will now refer
to
this as our ISP.

Our BM server is setup with a 2GB cache on Traditional storage with
suballocation off and an 8Kb block size. Some but not all of the other

performance tweaks have been applied.

Due to the very restrictive nature of the ISP setup the only DNS
server I
can contact is at the ISP and so I deleted all the root servers and
created a single entry under the Root Servers section of DNS to point
to
that.
I also have to set the option in BM HTTP Proxy | Cache Hierarchy
Client
to "Enable Cache Hierarchy Client" and also "Must Only Forward Through

Hierarchy" of type "Parent" and give it the address and port of the
ISP
proxy.

Given all the above the problems we are experiencing are:
1. 30-40 alerts a day stating "ICP Parent(s) - Not Reachable" (4 in
the
time it took to type this)
2. Many 502 Malformed reply from origin server and 504 Bad Gateway
errors
throughout the course of the day but particularly around lunchtime
when
surfing is almost impossible.
3. Variable performance throughtout the course of the day (I know
there
are a huge number of factors involved here) but very rarely is the
Internet experience comparable to ADSL even though we reportedly have
a
5Mb/s connection.
4. NETMON.NLM shows no transmitted or received errors but does show
that
we tend to get traffic only every 3 seconds with very little in
between.

Unfortunately the politics of the situation don't allow us to simply
move
elsewhere.

So - Can anybody see anything wrong with our internal setup (routing,
DHCP
etc), know of any issues between NetWare and CISCO switch/routing
gear,
have any suggestions as to how I might be able to narrow this down a
bit
or need further information?

MANY thanks for reading.