ARIN-PPML Message

[arin-ppml] debunking the myth that Moore's law helps

Michel Py wrote:
> There were several mentions of Moore's law lately. And with them, some
> implied comments that scalability issues would eventually be solved by
> it (assuming that it keeps working, which it has so far).
> 
> This is just not true, and here's why: yes, Moore's law is still in
> effect. But guess what: link speed / bandwidth/ TE requirements / etc
> have evolved also, possibly even faster than Moore's law.
> 
> Not so long ago, a 7500 was hot stuff BFR. A T1 was fast, a DS-3 was
> smoking, and an OC-12 was something only a telco would have.
> 
> Today, a T1 is slow; most consumer broadband offerings fitting a family
> budget are better. I have T1-speeds on my cell phone, and the CPU that
> runs said phone is probably more powerful than anything that ran a 7500.
> A DS-3 is not smoking anymore either, as FTTH and metro Ethernet have
> brought equivalent speeds (in urban areas) to the masses for a few
> hundred bucks a months. As of OC-12, it is a bit out of style too, when
> the typical FTTH home link is GigE and the name of the game in any data
> room is 10GigE.
> 
> So yes, Moore's law is still working. Does it help? Not much, as the
> operating requirements tend to be a moving target that adapts to
> whatever is available on the market.
> 

While I'm of the camp that Moore's Law will eventually taper off
(how are we going to produce transistors smaller than a silicon
molecule?) I have to point out that Moore's Law absolutely would
help, it just depends on it's implementation by the router vendors.

Moore's Law says nothing about actual power, the performance gain
that has accompanied transistor count doubling has happened first
due to clock speed increases, and now that they have reached the
limit of that, it's happening due to increasing parallelization within
the chip.  Moore himself states that it has nothing to do with
performance:

ftp://download.intel.com/museum/Moores_Law/Video-Transcripts/Excepts_A_Conversation_with_Gordon_Moore.pdf

How this applies to routing should be pretty obvious.  If you have
silicon in a router that has only a single internal path for packets
to travel, then your not going to be able to pass as many packets from
interface to interface as a router that has multiple, parallel,
internal packet paths - all other things being equal between the
routers.  More transistors, more parallel paths.  This is why the Cisco 
CRS-1 uses "massively parallel processing on the chip for flexible 
service delivery"

http://www.cisco.com/en/US/prod/collateral/routers/ps5763/prod_brochure0900aecd800f8118.pdf 
  (page 9)

Internet routing works the same way in the macro scale.  You wanna pass
a gig of data between point A and point B?  Well you can get a single
gig link or you can get 10 100mb links, you just need 10 routers to do it.

The problem I think is that the high power routers are so incredibly
expensive and multiple circuits are more expensive that network 
administrators are still thinking more in terms of getting 1 big hulking 
router and a high speed link rather than multiple smaller routers and
multiple links, when they need more throughput.  It simplifies network
design to think like this because you then don't have to deal with the
ickyness of load balancing.

But, Moore had an answer to that, because the other part of Moores law
has to do with cost - meaning that if you keep the same transistor count
on a chip, the chip cost falls.

I suspect if the Cisco CRS-1 was selling for $500 USD, rather than $500K 
USD, the answer would be pretty obvious to most people.

So yeah, I think Moore's Law applies.  I also think that the router 
vendors believe that most backbone providers have deep pockets and will
continue to pay lots of money for high power routers, thus they are
not competing on price for the high-end routers, and thus not exercising
that aspect of Moore's law.  Otherwise you would see yearly price drops
on CRS-1 line cards.  Also the carriers themselves have a vested
interest in keeping high speed link prices high, too.  They would
rather charge the same money every year for an increasing amount of
fiber capacity than lower their price every year for the same amount
of fiber capacity.

Whether this situation will continue is anyone's guess.  It definitely 
has not continued in the low end SOHO router and switch market.  There, 
everything is commodity, now.

Ted