A router is required to be connected between the 2 separate redundant
networks and another network that is non-redundant.
That description of your topology, reiterated later in your message,
is fairly generic. Especially beware that switched Ethernet networks
can employ *internal* redundancy: they can contain physical loops
for redundancy and in that case have to be managed using the
"spanning tree protocol" (STP).
If you want IP-based redundancy over Ethernet, you'll need dynamic
IP routing. A static routing table won't cut it.
1. Is it possible to connect a router to redundant ethernet networks?
ie. 1 port on the router connect to the primary network and the other
connected to the secondary network.
Definitely. Usually the scenario involves two point-to-point links
based on MAN Ethernet, providing geographically redundant
internetwork connectivity (perhaps to a non-redundant LAN).
Or do you suggest to use twin multipoint (multi-exit) Ethernet
networks? Not a problem with that, either.
2. Will it cause a loop in the network? (not very sure about this, but
it is something like connecting a network in circle.)
No it won't. Read something about the layered network model.
An IP router is an L3 device, so it won't cause a loop in your
Ethernet (L2) network.
http://en.wikipedia.org/wiki/OSI_model
3. Any cheapest product that will meet this requirement?
A second-hand Pentium PC with three Ethernet boards,
running Linux or BSD - worth a total of about $100 (maybe less).
A 486 would do the job too, as long as you can stuff three
NIC's onto its busses and make it work despite shortage
of ISA resources and PCI bus-mastering problems.
That's the lower end. At the higher end, you could consider
some Cisco routers (2600 and above, if memory serves)
or Catalyst L3 switches - the basic configuration
here could be worth about $3000 and there's no
upper bound.
Inbetween there's a number of alternatives based on
PC hardware and Linux - the routing performance
depends largely on CPU horsepower and bus bandwidth.
A modern desktop PC worth $500 can be a hundred times
faster than an elderly 486.
For Linux-based routers and the like, there are special PC-based
hardware platforms in a small form factor, e.g. with 2-4 onboard
Ethernet ports and a CompactFlash disk. Some of the common
brands are Soekris or Advantech.
The hardware cost is perhaps not the point though.
If you plan to set up a redundant topology using L3 (IP)
redundancy over Ethernet, you'll have to employ dynamic
routing - perhaps using RIP2 or OSPF. These are available
both on a Cisco and in the free unices - but you'll need some
additional knowledge to configure and maintain the boxes you
buy, be it Cisco or Linux.
http://www.cisco.com/univercd/cc/td/doc/cisintwk/ito_doc/rip.htm
http://www.cisco.com/univercd/cc/td/doc/cisintwk/ito_doc/ospf.htm
http://www.cisco.com/warp/public/104/1.html
http://www.cisco.com/univercd/cc/td/doc/product/software/ios120/12cgcr/np1_c/1cprt1/1crip.htm
http://www.geocities.com/Heartland/4394/work/ospf.html
http://www.opalsoft.net/qos/OSPF-14.htm
Any machine that needs to be dual-ported (for redundancy) must
run a dynamic routing demon in some form. "Routed" or "zebra"
come to mind in the Linux world - it appears that there is a Windows
equivalent, too.
Again, are we speaking dual-ported backbone routers or dual-ported
workstations? Imagine configuring dynamic routing on every windows
workstation... that's hell on earth
As a side note, for LAN redundancy, Cisco has a protocol called
HSRP, which is usually deployed in inherently redundant switched
Ethernet networks (using STP). I.e., there are no twin Ethernet
networks - maybe twin Ethernet switches. The client workstations
are single-ported (their individual Ethernet links are not redundant).
HSRP doesn't serve exactly the same purpose as redundant
L3 routing based on RIP or OSPF - HSRP is focused on providing
backup for failed machines, rather for network link failures (these
should be handled by STP).
There is a similar (open?) protocol called VRRP.
The configuration of HSRP/VRRP is not trivial, either.
Perhaps most importantly in your case, you may need several
Cisco switches and routers to create a fully redundant setup.
It is the recommended and more professional solution.
http://www.cisco.com/univercd/cc/td/doc/cisintwk/ics/cs009.htm
http://www.cisco.com/warp/public/cc/so/cuso/epso/entdes/hsrp_wp.htm
http://www.cisco.com/warp/public/473/62.shtml
(Unfortunately the best information is only available at paid Cisco training courses.)
You get what you pay for.
If you have time and are willing to learn, buy a PC box
and play with Linux. Otherwise, pay an expert to do it
for you...
For various reasons, some people prefer Cisco (or perhaps other
proprietary networking hardware vendors) while others prefer the
open-source DIY way. Traditionally, Cisco was the only solution
in the professional ISP arena, but many "retail ISP's" are nowadays
using PC+Linux for access gateways.
Frank Rysanek
