I have a system currently with a C7246VXR with a NPE-G1 and a MC28U. We have run into an issue with IP Address Spacing and want to reclaim with NAT. We have about 600 Modems on this system with about 400 of those Arris EMTA's. Normally our CPU usage is 10-15% at peak. When I have enabled NAT through a Single IP things seem to be ok, but as the day progressed the CPU usage kept creeping up. Even though our calls never peak over about 20 at any given time the cpu usage kept climbing. Its now at 86%.... Any suggestions?
--------------------------------------
Interface ActiveHiPriCalls ActiveAllCalls PostHiPriCallCMs RecentHiPriCMs
Cable3/0 0 7 0 0
Cable3/1 0 7 0 0
Total 0 14 0 0
-----------------------------------------------------
UBR7246VXR-1#sh ip nat statistics
Total active translations: 438 (0 static, 438 dynamic; 438 extended)
Outside interfaces:
FastEthernet2/0
Inside interfaces:
Bundle1
Hits: 40196868 Misses: 14508
Expired translations: 37885
Dynamic mappings:
-- Inside Source
[Id: 3] access-list 100 pool CPE refcount 2844
pool CPE: netmask 255.255.255.252
----------------------------------------------------------------
CPU utilization for five seconds: 99%/85%; one minute: 92%; five minutes: 84%
PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
88 2762356 4999626 552 9.62% 8.37% 7.48% 0 IP Input
173 362564 952971 380 1.57% 0.56% 0.44% 0 SNMP ENGINE
171 285744 538459 530 1.16% 0.53% 0.47% 0 IP SNMP
44 121756 1072414 113 0.49% 0.18% 0.13% 0 MCU BPE IP Enque
32 53364 30193 1767 0.16% 0.17% 0.16% 0 Per-Second Jobs
5 63084 5346 11800 0.00% 0.06% 0.05% 0 Check heaps
29 25900 17238 1502 0.16% 0.06% 0.07% 0 Net Background
21 31544 174113 181 0.00% 0.05% 0.07% 0 ARP Input
115 27908 37543 743 0.00% 0.04% 0.05% 0 DHCPD Receive
172 20784 213152 97 0.08% 0.02% 0.00% 0 PDU DISPATCHER
111 11584 47613 243 0.00% 0.02% 0.01% 0 CEF process
163 9480 510 18588 0.00% 0.01% 0.00% 0 Per-minute Jobs
159 1996 1025 1947 0.00% 0.01% 0.00% 2 Virtual Exec
38 5736 3065 1871 0.00% 0.00% 0.00% 0 CMTS CM MONITOR
---------------------------------------------------------------------------
Best to avoid NAT if at all possible.
What IOS version on the CMTS?
What addresses are you doing NAT for? CPE? MTA?
Are your MTA using private IPs?
Do all the MTA talk to a local SIP server/gateway (eg Asterisk?)
If you absolutely must do NAT, I've heard people say that using a Linux server in front of the CMTS will scale better.
It helps if the CMTS isn't the router doing the NAT translation, your gateway routers are typically better equipped to handle the load.
If you're pushing local IP's to the customer CPE, definitely offload that chore to your gateway router, not the CMTS.
Yes the NAT load was too much for the CMTS. I created a route-map for the private IP's for the MTA's and forward those packets out another interface to a dedicated NAT router running a Sup2. CMTS CPU load now avg about 20% and the NAT router only 2% :)