[60112] in North American Network Operators' Group
Re: WANTED: ISPs with DDoS defense solutions
daemon@ATHENA.MIT.EDU (Michael.Dillon@radianz.com)
Fri Aug 1 05:52:27 2003
To: nanog@merit.edu
From: Michael.Dillon@radianz.com
Date: Fri, 1 Aug 2003 10:51:21 +0100
Errors-To: owner-nanog-outgoing@merit.edu
>However, I would like to see Java or Other Language to run on the=20
routers,
>(I know you can install and play Quake on one vendor=B4s boxes) but I mean
>to do things really belonging to the router but so far I have yet to see=20
a vendor
>to take programmable boxen (outside their own development) seriously.
>Maybe it=B4s just too hard.
It's not hard at all technically. The hard part is that manufacturers do=20
not want to have to provide support for such flexible boxes when they have =
such a hard time just dealing with the complexity of existing general=20
purpose routers. Current routers are trying to be all things to all=20
people. They have far too many knobs and their routing software runs on=20
far too many hardware platforms. If they let you run scripts on the box=20
itself, there would be a combinatorial explosion in complexity which would =
make it a lot harder for the manufacturer's router gurus to help you.
But you can still build this yourself for some environments using=20
something like Click, the modular software router.
http://www.pdos.lcs.mit.edu/click/
You may not be able to get the type of interfaces that you need because of =
the need for driver support. Click simplifies the task by using Linux=20
device drivers but Click is not Linux and does not use the Linux kernel.=20
However you can integrate Click into a Linux system as a kernel module=20
rather like building a real-time Linux system
There are various manufacturers that make PCI WAN interface cards that can =
be used in such boxes.
http://www.pt.com/products/prodgroup=5Faccess.html
http://www.brooktrout.com/products/netaccess=5Fpri=5Fpci/
http://www.iphase.com/products/specificProduct.cfm/2/8/190
The real question is whether or not this type of build-your-own router is=20
feasible financially because of the small quantities involved compared to=20
a box manufacturer. And where in your network could you confidently deploy =
such boxes when you know that you will have to support them yourself.=20
Personally, I think the most interesting spot to try these boxen would be=20
at the provider edge and the customer edge because it's not hard to think=20
up value-added services that could be provided by such boxes and generate=20
additional revenue.
--Michael Dillon
