List members;
I have been reading the digest for some time now and I would like to ask some conceptual questions. I am a telecommunications systems designer and not a software guy, but I have tried and used a number of the tools availabel in the LARTC and Iptables, and many of the other great things included in Linux. We are designing commercial products and I don't want to ask my software people to do things that are not realistic or might cause instability in systems.
1) Is it feasible to use both dynamic and static routing systems in a Linux router? That is, if I am using the facilities in IProute2 and Iptables to do fair queing and bandwidth control, for selected subnets, will that work ok with one of the router daemons like ripd or ospfd? Will their be a problem with routing tables being confused between the two? I guess what I am really asking is what is the relationship between Ingress / Egress ques, Pre-routing / Post-routing tables and the daemons in the data flow within the box.
2) Given the above, are there any issues as far as throughput and processing horse power, are concerned? We are building on a 2.6 kernel and will be using PIV 2.4 to 2.8 GHz hyperthreaded systems with large amounts of RAM, but there will be other stuff running on the same machine, such as Squid cache server. Also, my satellite modulator and demodulators are on the PCI bus and will take some processor. We will be handling data rates of up to 8 mbps on the upstream (bits not bytes) and up to 40 mbps on the downstream. Hub routers may be receiving up to 1 Gbps of incoming data from remotes. These rates are maximum and not the norm, but rather the exception. Part of the reason for wanting to use OSPF for example is that this system is multi-destination and there will be new alternate routes available as well multiple routes appearing and disappearing.
3) I have read questions many times, from people who ask about ingress control (policing) and the same answer is always given, about the fact that you can't tell the internet to send data slower. Has anyone used the ECN congestion notification bits in the IP header? At least for an internal (read that edge) network, it should work just as well as the BECN controls in frame relay. This would allow the network to slow down the sender without dropping packets, and should work as well in UDP as in TCP.
4) Has anyone tried an accelerator for TCP with IProute2? I guess it would have to be behind the routing machine (LAN side), right?
5) Has anyone out there, worked with the HDLC driver in the kernel? It is built in and we have used it, but I don't know how to relate it to the Iproute2 commands. Is it as simple as setting up the same as you would for the zebra router?
Lastly, I wold like to thank all of the people in the open surce community and especially the folks that maintain this list for their efforts and their dedication to this project. You are appreciated, even if it isn't said often enough.
Loren Wells
CTO
Linksat America
Do you Yahoo!?
Yahoo! Mail - 50x more storage than other providers!
