From: James (james@towardex.com)
Date: Wed Sep 15 2004 - 23:09:10 GMT-3
On Wed, Sep 15, 2004 at 04:56:30PM -0700, Joseph D. Phillips wrote:
> I was wondering what current ISP practice is regarding hot-potato vs.
> cold-potato routing.
Depends on the problem an ISP is specifically looking to solve.
However, industry experiences and realities have shown that cold potato
in general causes more engineering nightmare than default hot potato.
Backbones who usually cold-potato their outbound traffic (egress) stand
ample amount of money increasing their transport capacity to remote
peering sites and POP's. Whereas, backbones who hot-potato their egress
traffic spend less amount of money on transports, but spend more money
increasing diversity in peering capacity as well as any transit relations
they have (e.g. settlement-free transit is one of the new ones lately,
or paid transit if the carrier is mix of tier2).
In my personal opinion, I find cold potato for egress traffic pretty
useless. I want my traffic to leave the network as soon as geographically
possible in route engineering. Networks today face situations where
often times, transit bandwidth costs are getting low as dirt sand where
transport prices aren't keeping up w/ the pace at the same market speed.
If it were you, would you rather spend more money having diverse connections
with other ASN's, than increasing capacity of your own network transports?
Granted, you must ensure you do proper capacity planning of your transports,
but I prefer diversity more when it comes to traffic engineering.
>
> According to one article I read, ISPs sort of go right down the middle
> -- not trying to get rid of traffic right away and not trying to do too
> much QoS on it, either.
Although not defined anywhere by the industry often, some people and myself
like to call this "meshed-potato" where it is between hot and cold --
We use hot potato to egress the traffic as fast as geographically possible
to our transit carriers. At the same time, we use cold potato to ensure that
return traffic, arriving back to our network, remains on our transit provider's
backbone as long as possible, instead of having them offload it to our network
in hot potato style.
So while we send our outbound traffic as fast as we can off our network to
our transit upstreams, since we are paying transit carriers to move bits anyway,
and they gaurantee their network performance with SLA, we makes sense for us
to use their network as long as possible for traffic coming back to a particular
point in our network.
Then again, if this is against a peering partner, instead of transit carrier,
then meshed-potato is something you should *never* do until you ask your partner
first... or you may end up facing consequences up to and including unexpected
de-peering :)
-J
-- James Jun TowardEX Technologies, Inc. Technical Lead Network Design, Consulting, IT Outsourcing james@towardex.com Boston-based Colocation & Bandwidth Services cell: 1(978)-394-2867 web: http://www.towardex.com , noc: www.twdx.net
This archive was generated by hypermail 2.1.4 : Fri Oct 01 2004 - 15:00:44 GMT-3