ASR9001 PPPoE New sessions per second

[dfranjoso]

Hello All,

Does anyone know the amount of new PPPoE sessions can the ASR9001 handle per second? Thanks!

David

[xthuijs · Accepted solution]

never an easy question to answer... same as asking for the top speed of a car... wind in the back? downhil/upslope? who's driving it, got luggage and your family in the car or alone... things like that...

Same applies to CPS.

Theoretically XR BNG can do 200 cps.

but the minute you add features that require NP programming such as ACL and QOS that cps might drop.

If using external radius, different ball game all together; does it support extended source ports (so support more then 256 outstanding requests). Is accounting on (and to the same server) and is the accounting database local to the server or an external ODBM/SQL connection or what...

I have always heard people stating that their radius is never the bottleneck, until you show the difference in perforamnce with local authentication.

Now XR doesnt do local auth, so it comes down to no auth on the dyn tpl.

This is what I call race track, no auth and no features other then just enablement of sessions.

When you then add acl or qos and then acl and qos, which are the main features

and maybe add (interim) accounting later on, you can nicely baseline what the numbers are.

On average I would say that call rates in between 50-100 with radius and features is "industry average" and most likely to expect. Yes wide range, but I want to be realistic.

does that help?

regards

xander

[dfranjoso]

Thanks Xander. I need to have a baseline for my testing. Has i said to you in Cisco Live, the work you develop here in this forum fills the Gap between the Cisco crappy Documentation and the information that customers actually need to plan and configure their networks.

[xthuijs]

Thank you very much david! It was also very nice to meet you in person! Glad we had that opportunity!

The support forums is part of the XR strategy to bridge the gap between cco docu and what we can really use as from engineers for engineers. So you'll likely see more participation from other 9k eng's also coming soon with more technotes etc. At the same time the cco doc improvement is also an official project, but that might not go as fast as I would want.

Talk to you soon!

regards

xander

[j.restaino]

Hi Xander,

We have ASR9010 BNG, should we expect the same limit? We had a problem last week. We sustained several thousand new pppoe session requests  per second for a least an hour, and then we suffered a crash in the radiusd daemon. We try to restart several times the radiusd, but despite of appear to be up, the BNG do not send radius messages. Finally we had to reboot the BNG.

Is it possible that this limit affects the radiusd in this way?

Regards
                      José

[xthuijs]

Hi Jose,

RSP2 is a dual core PPC, RSP440 is a quad core Intel, 9001 has a quad core PPC.

So while the RSP440 with its intel is a bit more powerful then the 9001's QC PPC, they are very close. What I mean by that is yes there is a difference, but it is small.

Although the problem statement you raise is rather clear :), I am missing the detail to make a thorough assessment as to why this happened. BNG in A9K has flow control by default, so the number of incoming requests should be streamlined and of course process crashes should never happen. Depending on the release of sw you are running currently, there may be a bit more tricks to do to help mitigate and handle this situation. I would want to recommend opening a new discussion for that or a TAC case so we can analyze and evaluate the different traces that were left of this problem to give you the right advice.

cheers!

xander

[j.restaino]

Thanks Xander,

We open a new discussion thread "https://supportforums.cisco.com/discussion/12386091/asr9010-bng-radiusd-crash".

Thanks for your help.

Regards
                  José