• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Heavy linux router reccomendation

A

acemcmac

Lifer
I am trying to build a proof-of-concept router using Devil-linux for a theotretical 600 node network. Protocol and NIC issues asside, how heavy do I need the hardware on this to be? I am currently leaning towards a Barton 2500 on retail heatsink and 256 megs of pc2700 on a BIOSTAR KM400 M7VIZ. Overkill? Undercooked? any ideas?

Instability is not an option. This device is a proof of concept for a switching fabric load alleviation system and has to be able to take punishment without dropping a noticable number of meaningful packets.

thanks for the input. :beer:
 
I don't know what "switching fabric load alleviation" means.

How fast of a network are you running? How many ethernet ports (or other type of networking ports) are you running?


A 133 pentium is enough to route between 2 networks at full 100mb/s speeds. I suppose a 1ghz with 128 ram has the ability to take anything you can through at it, thats probably much more powerfull then any router you'd find anywere.

Quagga is the routing software for Linux. Linux has mostly built-in bridging capabilities and has support for nice stuff like spanning tree protocols, it has the ability to do stuff like NAT/masquaranding and to forward packets around. I suppose you could set up a bunch of nic cards together as a mirror'd bridge or something and get a nice switch, although I never thought about doing that before...

Quagga has the ability to create a full-blown linux router and has support for such things as RIP, OSPF v3/v4, and border gateway protocols. (quagga is a fork off of Zebra, which was the original name, but last it seemed not be developed much. So Quagga people made a fork of it.)
 
Peak throughput?
How many rules?
How big of a state table?
Looked into high availability failover?
Budget?

I'd personally go with a couple of boxes in a HA failover. Either late durons (1.8?) or bartons. And probably as much ram as I can get a hold of.

EDIT: I was thinking firewall for some reason. Ignore the firewall related questions. 😛
 
What is this switching load fabric stuff? Like a network that runs a large network's backbone or something? ATM/Frame relay stuff? Would the PCI buss in the computer be a limiting factor?
 
Originally posted by: drag
What is this switching load fabric stuff? Like a network that runs a large network's backbone or something? ATM/Frame relay stuff? Would the PCI buss in the computer be a limiting factor?
Click to expand...

You know. Alleviating the load from the switching fabric. 😕
 
In my spare time over the last few weeks, I have been trying to develop a proof-of-concept for a linux box that can be placed inline either between the distribution and access layer, or within the access layer that can drop "white noise" packets while letting all other traffic pass through. This is a MAC level concept, no DCHP or NAT involved... More than 85% of this lan's traffic is UDP broadcast of p2p directories, nachi-b and crap like that. My discussions with other memebers in the networking forum have led me to theorize that the switching fabric is overloaded (3-4 years old) and this sudden, massive boost in meaningless traffic is causing the massive service blackouts the lan is having. Deployment issues aside, I am trying to develop a linux box that can handle extreme loads, dropping the appropriate traffic without compounding the problem for meaningful traffic. Simply wondering how much horsepower would be required to make that an iron-clad gaurentee... current plans are for the onboard nic and an intel pci nic, but will probably later be replaced by fiber pci nics....

Did that clarify anything?

lots of elbow grease and almost no money available... switch replacement is not an option and the fabric capacity is not scalable. over 800 nodes will be added to the already saturated fabric within the next 6 months.... looking at 13-15 deployments of these boxes across 4200-4500 nodes... what a mess... I've gotta get something working here lol 🙂

Edit: will look into Quagga thanks for the tip 🙂
 
You must log in or register to reply here.
Back
Top