Quantum: Above 1K ports, every port create takes more than 1+ min

[puppet-py]

in a scale setup(COI based), comprising 64 compute nodes and 1 controller node, every port create takes 1-2secs until it reaches a count of 1000. Since then every port creates takes 1+ min, it does NOT involve VM create, just port create.

The above 1000 ports are and subsequent port creation are being done in one network-segment and using dhcp policy-profile of N1K VSM

Observation: On shutting down services like nova(all components), cinder, glance, the port-create time got exceedingly improved and it was taking only 1-2 secs and I could go upto 4K ports.

Using the below REPO:

root@bldsrv141-n1k:/etc/puppet/manifests#cat /root/cisco-grizzly-manifests/manifests/puppet_modules.py | grep REPO_NAME
# Set REPO_NAME below to the name of the yum or apt repository you want to use.
#REPO_NAME = "grizzly-proposed"
REPO_NAME = "grizzly/snapshots/2013-10-09"

Please Note: Filing this bug as a placeholder as we will add more logs and data on our next repro-cycle