IP VPNs and Reliability

We've been researching various VPN options for past year or so and we're
in discussions with yet another vendor regarding their MPLS-based
solution. All but one of the vendors we've talked to suffer from a
similar 'problem', but I'm curious to get your opinions on this. I'm
wondering if I'm more concerned about this problem than I should be.

Most of the vendors we've talked to only have a single POP in our area.
It seems to me that if we were to choose an option like that we'd, in
effect, have a network of point-to-point circuits terminating at that
one POP. I'm not that familiar with the hardware but I suppose it's
possible that all of those circuits would be terminating on just a
couple of huge routers. My concern is that a problem in one POP could
take down our entire network, whereas with our current frame relay
vendor we have great geographic diversity with the frame/ATM switches.

The vendors who have a single POP that we've looked at are MCI and
Virtela. [I must say hello to Gene and Brice at Virtela since I know
they'll get a copy of this. <g>] Given that these vendors seem to have
rather robust infrastructure, am I more worried than I need to be that
all of our connections might terminate at the same spot? I'm sure those
POPs are almost bullet-proof, but I'm still a little leary to put all of
our eggs into a single basket like that.

Any opinions would be welcome, especially if my assumptions are wrong
and you'd like to correct me.


Why don't use 2 vendors in every _important_ site, using routing (OSPF or
EGRP) for VPN balancing? Or just add 1 low performance link to _another_
vendor and build traditional (tunnel + IPsec) VPN structure in addition to
MPLS-based provider's solution?