[afnog] Connectivity to www.afrinic.net-BGP qtn

[Mark Tinka]

On Thursday 08 July 2010 06:00:10 pm Walubengo J wrote:

> ...and so just thinking in terms of engineering the
>  automatic backups systems for IP (or whatever) traffic
>  btwn TEAMs and SEACOM, would i be right in presuming the
>  following?
> 
> 1. that the Layer1 Circuitry btwn SEACOM and TEAMS should
>  be in place & then 2. that ISP/IBP connecting into both
>  fibers (Layer2?) then have to enter an interconnection
>  agreement as well as connect IP routers btwn them?
> 
> walu.
> nb: still think step 1 not necessary though since step2
>  alone could provide the redundancy sort...

The bottom line is this:

	- Submarine operators along the same path normally have no
	  incentive to interconnect with each other on the on-set,
	  as they are competitors.

	- Those that interconnect would do so at an aggregate
	  level, i.e., between SEACOM and TEAMS, not between
	  SEACOM's customers and TEAMS' customers, which means just
	  because there is an interconnect, doesn't mean you will
	  enjoy it when there is a failure on your favorite cable
	  system.

	- Circuit-switched networks are not very efficient, as
	  paths need to be pre-provisioned before they can be used,
	  and remain that way regardless of whether you use them or
	  not. So even if there's an interconnect between SEACOM
	  and TEAMS, it might not have anything to do with the path
	  you're interested in at the time of failure.

	- Multiple cable operators are NOT obligated to
	  interconnect with each other (again, especially if
	  they're direct competitors in a given market).

	- Automated switchovers at the Layer 1 level have been
	  spec'd, but haven't really worked out reliably in the
	  real world, e.g., APS (Automatic Protection Switching),
	  a.k.a 1+1, involves buying one "live" path and one
	  "protect" path on a single cable system. If the live path
	  fails, APS switches traffic over to the protect path.
	  Although this mechanism is well-documented, it is always
	  cheaper to buy one linear (linear = unprotected) service
	  from Provider A and another linear service from Provider
	  B and implement your own redundancy at the IP layer using
	  your IGP and BGP. No point in paying for APS, as you're
	  losing money on a circuit on which you can't put
	  bandwidth half the time.

	- Cable operators generally have no interest in the
	  payload, so they won't build their networks around what
	  you plan to run on them. The issue of application
	  resiliency is left up to the user.

	- ISP's are fully responsible for mitigating their risk re:
	  circuit failure. Have diverse paths, have sufficient
	  bandwidth along those paths. Don't rely on the Layer 1
	  provider for redundancy.

Hope this helps.

Cheers,

Mark.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
URL: <http://afnog.org/pipermail/afnog/attachments/20100708/a8537865/attachment.pgp>