Showing results for 
Search instead for 
Did you mean: 

SCE and users with multiple IPs

I am testing SCE2020's with an MGSCP setup and have found that if a user has more than 1 IP (which many do) then their traffic is likely to be split across different devices which means the RDR's for the user are sent from multiple SCE which totally breaks the users quota.

What is the best solution for this problem? We need MGSCP for the capacity but also need to get the correct quota data from the subscriber manager for billing purposes.

Are there plans for the Subscriber manager to support receiving subscriber RDR's from multiple SCE?  Is there a reason it does not support that as the RDR's include the sending SCE IP, so if RDR's were being sent from 2 SCE then it should be able to distinguish the 2 sets of data rather than mixing the 2 seperate counters and calculate accordingly?



Shelley Bhalla

Interesting use case!

The load balancing on the MGSCP is done on the ip address (source or destination) causing different ip addresses to take different paths. There is no alternative to this unless you NAT these ip's elsewhere so that the SCE see's only 1 IP from this user.

For SM question, SM does support multiple SCE's at this time. Please review the latest user guide for the SM.



>Interesting use case!

But surely there are many providers that have users with multiple IP's and need the bandwidth of an MGSCP setup? We cant yet justify the cost of throwing away SCE2000 and going 10G and SCE8000.

>For SM question, SM does support multiple SCE's at this time. Please review the latest user guide for the SM.

The SM does support multiple SCE, but not receiving RDR updates for a single user from multiple SCE simultaneously as would be the case of a subscriber with 2 IP's routing via different SCE in the MGSCP topology - unless I am mistaken?  I cant see mention of that in the docs and in practice the bucket counters get muddled resulting in corrupt usage data.


You are correct. SM does support multiple SCE and ip ranges but the built in safety mechanism of seeing 1 subscriber only on 1 SCE will not be changed.

In my term of supporting this product for 4 years, I have not come across a "working" scenario where 2 seperate IP's were used for a subscriber in an environment where we have multiple SCE's in a MGSCP  setup so I understand the limitation you see in your environment.

In scenarios where there were multiple IP's given to a subscriber it was a range of ip's given to the subscriber and the "support_ip_ranges" parameter in the p3sm.cfg file was changed to "yes". This allows multiple IP's to be assigned to the subscriber but they all have to traverse a single SCE box to avoid an issue on the SM.

Given that the behavior will not change, The only option we have is to either do billing for each ip separately or NAT the subscriber ip's before it hits the SCE.



Cisco TAC

Fabian Prosman
Cisco Employee

Hello Jim,

There is one thing that pops into my mind reading your description. Is this information also relevant at the NAS side? Who are those subscribers, companies (range) or home users (expected single IP)?

I don't want to make any assumption from the network topology but it may be that 2 different people shares the same account, then having the 2 IPs.

Would you have an example to share?

I'm also wondering whether you could overcome this situation with the SM API or not. I've never used it personally but it might be worth having a look at the possibilities.



Tom Debruyne
Cisco Employee

Hi Jim,

Depending on your setup, you could maybe consider distributing the load among the SCEs with policy base routing instead of port channel load balancing.  In this case, you could group the IPs by access list to chose through which SCE they will be sent.  That solution could work as long as you can create access lists in such a way that 2 different IP address from the same subscriber will be sent through the same SCE.



Hi Guys,

In the scenario as above, when there is a subscriber with multiple IP's and the SM downloads the packages to the SCE. when running an MGSCP setup across 2 SCE's, do the SCE's talk to each other (directly or through the SM or CM) in order to provide a single policy across both boxes or is the policy on each SCE enforced independently?

In this scenario will a subscriber with a 100mb/s policy get 200mb/s in total because his different IP's are being load balanced based on source IP across the two SCE's or is this problem just related to the way the information is being reported to the SM for billing purposes.



In MGSCP the SCE's dont talk to each other but they both interact with the SM and CM independently. The policy of the subscriber is the same, It is dictated by the SM when the Login Event is sent from the SCE for that subscriber.

When we see a subscriber log in on more than 1 SCE, the SM will trow an error message and if the issue goes unnoticed we can go as far as quarantine the SCE on the SM.

In such a scenario when a subscriber has a group of IP addresses attached to his profile. No matter where he logs in (as long as he logs in on only 1 SCE) he will be given the same policy of 100Mbps. not 200Mbps because he logged in 2 times on 2 different SCE. He wont be allowed to log in 2 times anyways.

The SM (for billings and sanity check) will only allow a user/subscriber to be logged in only once in the system. By system I mean all SCE's in or outside of MGSCP sending login events to the same SM.