Cisco Cisco IOS Software Release 12.4(2)XB

Only five instances of the image can be loaded onto the MWAM.

The same Cisco IOS image must be loaded onto all processor complexes on the MWAM.

Session console is provided by TCP connection from the Supervisor module (no direct console).

Available memory for bootflash for saving crash information files is 500 KB.

Only five files can be stored in the bootflash filesystem.

To avoid issues with high CPU usage, we recommend the following configurations:

To reduce the CPU usage during bootup, disable logging to the console terminal by configuring 
the no logging console global configuration command.

To ensure that the HRSP interface does not declare itself active until it is ready to process a 
peer’s Hello packets, configure the delay period before the initialization of HSRP groups with 
the standby delay minimum 100 reload 100 interface configuration command under the 
HRSP interface.

To minimize issues with high CPU usage for additional reasons, such as periods of high PPP 
PDP processing (creating and deleting), disable the notification of interface data link status 
changes on all virtual template interfaces of the GGSN using the no logging event link-status 
interface configuration command.

! 

interface Virtual-Template1 

description GGSN-VT 

ip unnumbered Loopback0 

encapsulation gtp 

gprs access-point-list gprs 

end 

For implementation of a service-aware GGSN (GGSN Release 5.2 and later), the following additional 
important notes, limitations, and restrictions apply:

RADIUS accounting must be enabled between the CSG and GGSN to populate the Known User 
Entries Table (KUT) entries with the PDP context user information.

The CSG must be configured with the quota server (QS) addresses of all the GGSN instances.

Service IDs on the CSG are configured as numeric strings that match the category IDs on the 
Diameter Credit Control Application (DCCA) server.

If RADIUS is not being used, the Cisco CSG is configured as a RADIUS endpoint on the GGSN.

On the SGSN, the values configured for the number GTP N3 requests and T3 retransmissions must 
be larger than the sum of all possible server timers (RADIUS, DCCA, and CSG). 

Specifically the SGSN N3*T3 must be greater than:

2 x RADIUS timeout + N x DCCA timeout + CSG timeout

where:

2 is for both authentication and accounting.

N is for the number of diameter servers configured in the server group.