(05-13-2020, 02:46 AM)ghost180sx Wrote: How easy is it to setup bridging? Is it a very simple option that you pass through the /proc interface or using something like ifconfig or route?
Setting up Linux as a router is easy:
Code:
sysctl -w net.ipv4.ip_forward=1
But your favorite distribution probably has a file where you can add this, e.g. /etc/sysctl.conf in Debian
Let's assume your LAN uses IP range 192.168.1.x and the FDDI systems use 192.168.2.x. The FDDI router box has one IP in the LAN range (let's say 192.168.1.10) and one in the FDDI range (192.168.2.1). There's a LAN-to-internet router of course (192.168.1.1). This router offers a default route (to internet) and a static route for the 192.168.2.x range via 192.168.1.10 (the FDDI router). That way all systems on my LAN can transparently reach the FDDI range. I could configure the FDDI route in every system in my LAN and save the traffic to/from my LAN router but decided against it (flexibility and easy of use vs. traffic).
The tricky part is SGI systems with both an ethernet and an FDDI interface. You end up with two routes to these systems: one via 10Mbit ethernet and one via 100Mbit FDDI. In the past I had my SGI systems in my LAN, so the fastest way from a system on LAN (let's say 192.168.1.30) to Onyx (let's say 192.168.1.20 & 192.168.2.20) would be via the 192.168.2.x network. No amount of DNS trickery, 'routed' magic and hop weights could convince anything to leave the 192.168.1.x network and go via 192.168.2.x instead.
I ended up disabling the 10Mb/s ethernet interface on these SGI systems (wired but disabled in IRIX), and these days my SGIs sit on their own IP range. The IP ranges by themselves are not the solution: the fastest way from e.g. my Origin 200 (Gbit ethernet) to my Crimson (10Mb/s ethernet, 100MB/s FDDI) is via FDDI, even though the ethernet interfaces are all in one IP range and FDDI in another.
If someone has a pointer to the recommend way to solve this I'd like to hear it. There's got to be a more intelligent solution than e.g. an IP range for every possible wire speed.
(05-13-2020, 02:46 AM)ghost180sx Wrote: Have you setup a ring topology?
My understanding was that if you have a single host in your ring topology that goes down, packets could be dropped. Is there a built in bypass to the PCI card to prevent that? Maybe I should just read books on the subject... I have them somewhere from my network study courses.
I keep a copy of the old
Practical Guide to FDDI on my server, which has most of the info you need to build an FDDI network.
You are right about the dual ring topology. The downside of a ring is that all systems are expected to be up and running -- loose more than one node and your ring is gone. Systems on the ring are connected with 'DAS' cards (dual attach, A & B ports). But FDDI also knows the concept of point-to-point connections ('SAS', single attach, M ports). An FDDI concentrator normally sits on the ring, so it's uplink is DAS, and it offers a number of M ports to connect stations. Effectively, a concentrator allows you to build a local tree topology which is attached to the ring. The stations attached to the M ports don't need to be powered up all the time, the concentrator takes care of that. It behaves a lot like an ethernet switch.
So, my FDDI topology looks like this:
Code:
LAN <-> FDDI router
/ \
|ring|
\ /
Concentrator
| | |
SGI SGI SGI
Technically I have a ring, but it's the most minimal ring possible with only two stations on it. This setup doesn't need bypass switches.
FDDI networking is well supported in IRIX, all the way down to the installation miniroot (but not the PROM). If I boot a miniroot on an old 4D series over 10Mb/s ethernet, it will detect FDDI interface and run the actual network installation using that.
NB: It just occurred to me that I might also have set up the FDDI router as a bridge, with FDDI and SGI ethernet in the same IP range. Setting up Linux as a bridge isn't too complicated, *but*
(1) The SGI ethernet interfaces are directly attached to the LAN router, the FDDI interfaces via the FDDI router. You'd have to set up a non-standard netmask on the LAN router to send part of the range via the FDDI router (bridge).
(2) FDDI uses a larger MTU than ethernet, so packets going FDDI -> ethernet need to be fragmented. Not sure how that works with a transparent bridge.
Ultimately, I don't care so much about whether multiple IP ranges are in use, I care that traffic automatically chooses the fastest route between two systems. I also have some vague plans to move the LAN to 802.1X Port-Based Network Authentication for improved security, at which time things need to be redesigned anyway.