🚀 go-pugleaf

zur Illustration fuer das p.t. Publikum : 

From: Geert Jan de Groot <GeertJan.deGroot@ripe.net>
X-Organization: RIPE Network Coordination Centre
To: "Nick Vermeulen" <Nick@wirehub.net>
Cc: local-ir@ripe.net
Message-Id: <9602011158.AA00635@ncc.ripe.net>
Date: Thu, 01 Feb 1996 12:58:41 +0100

On Thu, 1 Feb 1996 02:32:11 +0100  "Nick Vermeulen" wrote:
> imho, most multi-homed sites can do with a partial routing table; 
> organisations who want or need to have full routing tables should put 
> extra RAM in their backbonerouters to meet demand. When 64 MByte 
> isn't enough anymore...buy a cisco 7500 series which can hold 128 
> MByte RAM. you can minimize the number of your full-routingtable 
> routers with ebgp multihop features (for multi-homed customers)
> in the mean time...monitoring routing tables and notifying net-admins 
> that advertise less efficient aggregates will help.

If the problem is just adding a few SIMMs then people would do that.
The real problem is that current routing protocols require more growth
in resources such as CPU and memory than technology improvements
can offer. One can roughly say that CPU's double in speed every 24 months;
and as you know the Internet doubles every 10 months, thus the net
grows faster than new technology can offer additional resources.


The only way to cope with that is to make a 100% growth of the
net result in much less than a 100% growth in resource requirements;
the only mechanism of which we know it's working is hierarchical routing,
which requires people to renumber if they change provider.

If people don't renumber, then they will add significantly to the strain
the net is seeing already and the whole discussion started because one
of the US ISP's announced to implement some policy to archieve this
which has very bad side effects.

There are a couple of approaches to archive hierarchical routing:
- prefix-length filtering as announced by said US ISP
  (this is by far the worst choice IMHO)
- prefix-charging
  (this has some non-technical problems which aren't resolved yet)
- community consensus on limitation of prefixes
The whole issue has been discussed at the RIPE-meeting earlier
this week; I don't think I saw you there.

It isn't a matter of adding some SIMMs; some of the people who have
brought up their concern on this are running Cisco hardware that isn't
even available commercially yet, and yet need to do everything they
can to keep their boxes running. 
They are seeing some 50% of CPU utilisation on their routers during
normal operation, see the growth curve in this and know that
once it reaches 100% (causing the box to roll over and die), 
they won't be able to buy a faster box because none is available.

I have seen MCI slow-starting one of its routers by hand during the
IETF in Dallas some weeks ago and it was a very scary sight.

The technical background on the current problems has been 
discussed back and forth on the IETF CIDRD working group, and
I invite you to read the mailing list archives as your suggestion
and its problems has been discussed there already.

[...]

und

From: Per Gregers Bilse <bilse@eu.net>
Organization: EUnet Communications Services BV
Date: Thu, 1 Feb 1996 22:43:29 +0100
To: [internal tech-list]
Subject: MAE-East bumpy

We are currently seeing some bumpiness on MAE-East, looks like
LAN problems or maybe iBGP meltdown in some of the bigger networks.



Nicht dass jetzt Panik angesagt waere --- aber weit draussen ( besser : 
drinnen ) rumort's ...


-- 
<------------------------------------------------------------------------->
Georg Chytil                                    CG46  chytil@Austria.EU.net
EUnet EDV-Dienstleistungs-Gesellschaft m.b.H.   http://www.Austria.EU.net/
Fon : +43/Vienna/31376                          Fax : +43/Vienna/3106926