North American Network Operators Group

Date Prev | Date Next | Date Index | Thread Index | Author Index | Historical

Re: Scalability issues in the Internet routing system

  • From: Richard A Steenbergen
  • Date: Tue Oct 18 16:57:08 2005

On Tue, Oct 18, 2005 at 01:41:53PM -0400, vijay gill wrote:
> Moore's law for CPUs is kaput. Really, Moore's Law is more of an 
> observation, than a law. We need to stop fixating on Moore's law for the 
> love of god.  It doesn't exist in a vacuum, Components don't get on the 
> curve for free.  Each generation requires enormously more capital to 
> engineer the improved Si process, innovation, process, which only get 
> paid for by increasing demand.   If the demand slows down then the 
> investment won't be recovered and the cycle will stop, possibly before 
> the physics limits, depending on the amount of demand, amount of 
> investment required for the next turn etc.

Moore's "observation" would also seem to apply only to the highest end of 
components that are actually "available", not to what a particular vendor 
ships in a particular product. Of course we have the technology available 
to handle 1 million BGP routes or more if we really needed to. Processing 
BGP is fairly easy and linear, modern CPUs are cheap and almost absurdly 
powerful in comparison to the task at hand (they're made to run Windows 
code remember :P).  But if there is no reason to make a product, it 
doesn't get made.

Comparing consumer-grade PCs which are produced in the millions or 
billions with one small part of a high-end router which is produced in the 
thousands, and which is essentially an embedded product, is a non-starter.
The product that is sold does what it needs to do and no more. There is no 
reason to design a scalable route processor which can be easily upgraded. 
There is no need to ship the latest state of the art Dual Xeon 3.6GHz with 
support for 64GB of DRAM that will last you for your BGP needs for the 
next 10 years, or even to throw in 128MB of SSRAM for a few thousand bucks 

Everyone needs to sell their latest and greatest new product on a regular 
basis to stay in business, it is no different for router vendors. They 
sell you a $20,000 route processor that you could pick up from their 
vendor directly at Qty 1 for $1000, and the markup goes into the what you 
are really paying for (R&D, software dev, testing, support, sales, and the 
bottom line). A few years later when you need something a little faster, 
you can buy a new router. Besides, you probably needed to come back for 
the next-gen platform and interfaces anyways. Most customers are barely 
qualified to add RAM to their million dollar router, and I doubt the 
vendors see any need to change this.

> Also, no network I know is on the upgrade path at a velocity that they 
> are swapping out components in a 18 month window. Ideally, for an 
> economically viable network, you want to be on an upgrade cycle that 
> lags Moore's observation. Getting routers off your books is not an 18 
> month cycle, it is closer to 48 months or even in some cases 60 months.

Want to venture a guess as to how many networks are still operating 
routers with parts at the end of that 60 month cycle (purchased in 2000), 
which were in turn based off of 1997 technology back when they were 
manufactured in 1998-1999?

Richard A Steenbergen <[email protected]>
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)