Q: Sizes of Existing and Planned Fully Meshed IPSEC VPN (Tunnel Mode)

Mailman List Mirror (Read Only)
1

Lots of old and dear friends in NANOG to say HELLO!, it
has been many, many years since I've posted here. Hopefully
this very simply question will not be very controversial :slight_smile:

I did a search of the archives on VPN with keywords 'size' and
'mesh' only to find a thread debating the merits of MPLS. This
was an interested thread and some of my old and dear friends
were active in that discussion. Please allow me to ask a simple
less-technical question. My apologies if this has been discussed
and I missed it in the archives.

We have a Cisco IPSEC based VPN with over 110 edge routers
in a full tunnel-mode mesh, mostly 'big hunking routers' with
average CPU utilization under 15 percent. The VPN is
controlled by a single organization, under centralized admin.

Are there larger fully meshed VPNs out there in ISP land?

Are there any 'real-tangible issues' with a fully meshed VPN
at the size we are talking (around 120 sites fully meshed)?

The marketing hype tends to be great. I like -vadim's closing
comments in:

http://www.merit.edu/mail.archives/nanog/2001-08/msg00311.html

as follows:

"Sometimes older ways are simply better." --vadim

This seems to be true regarding a simple fully-meshed IPSEC VPN
in tunnel-mode, right NANOG geniuses? Is 110 fulled
meshed edge routers considered big??

Finest Regards, Tim

www.silkroad.com

2

We have a Cisco IPSEC based VPN with over 110 edge routers
in a full tunnel-mode mesh, mostly 'big hunking routers' with
average CPU utilization under 15 percent. The VPN is
controlled by a single organization, under centralized admin.

Are there larger fully meshed VPNs out there in ISP land?

Are there any 'real-tangible issues' with a fully meshed VPN
at the size we are talking (around 120 sites fully meshed)?

My god, your job is worse than mine :wink:

We have a fully meshed Cisco-VPN with half that many edge routers, and we
have more than 100 open bug reports with Cisco. Every single release they
have shipped has an issue that means we can't run it in one or more sites.

We're back to doing something I swore I would never do after working in the
NavSea MAN -- running the very latest code in brave but futile hope that
they've fixed something. 90% of the supposed 'bug fixes' they give us break
something else.

With 110 peers fully meshed, you must have only a single access-list
entry per site AND not all your sites talk at the same time. Until very
recently there was a hard cap on IPsec SAs that we kept slamming into
due to multiple access-list entries per site gives you (source+remote)^2
number of SAs...