[c-nsp] ASR1002 LNS scaling

[CIT - Chris Balmain]

Hi all,

We're trying to migrate some broadband users to an ASR1002, but seem to be having some problems relating to scalability for PPPoVPDN.

The session setup rate seems to be very poor - our configuration is somewhat complex, but we still found that the same numbers worked OK on other ASR1K platforms such as ASR1001 and ASR1001-X.

Our per-subscriber configuration includes:

*         ACL rules (about 10)

*         Class-default egress shaper

*         Terminating in VRF

We find that in production during the mass-reconnection of around 3,000 users, some subscribers simply wouldn't reconnect even after an hour or so.

In lab testing, we found that the session setup rate seemed to be OK with an ASR1001 acting as LNS (5-10mins to establish 5,000 PPPoVPDN sessions) - but with the ASR1002, it took several hours to do the same.

We've tried experimenting with Call Admission Control (CAC) but didn't find that to help either. Running latest 03.16.04b.S.

Anyone else experienced this, found a workaround?

Regards,
Chris