[rbak-nsp] B-RAS vendors

[Marcin Kuczera]

On 2013-07-15 18:20, Florian Lohoff wrote:
> On Sun, Jul 14, 2013 at 11:09:34AM -0300, Tomas Lynch wrote:
>>> Does anyone have experience of loading this many or more subs on an
>>> SE600/1200 with XCRP4?  We are not even using any QoS, NAT or any other
>>> features at present.
>> No, all I have seen was no more than 96K per box.
> With 96K sub and 100 Subscriber/s setup rate you'll need 16 Minutes to
> be back in service after service downtime for whatever reason. And
> thats under best conditions of your whole network.
>
> There is no good reason for these large systems although we have
> done that in the past with the SEs.
>
> The SE is doing fine - But suddenly you start thinking about RADIUS
> Server network latency and using multiple ports on client and server.
> (RADIUS has 256 slots for in flight packets per communication link)

That's important element you've just mentioned.
Radius - right after a crash it is really heavy loaded. At the moment we 
have ~15k subscribers (CLIPS/PPPoE, mixed)
and we decided to have Radius database kept on SSD.
Before in heavy load situations database was not able to respond quickly 
enough to radius (freeRadius, mySQL).
Now, with SSD it is much faster and all the recovery after eventual 
linecard crash is much faster.

Regards,
Marcin


>
> Flo
>
>
> _______________________________________________
> redback-nsp mailing list
> redback-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/redback-nsp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/redback-nsp/attachments/20130716/358121c3/attachment.html>