[c-nsp] Any BGP fine tuning recommendation while Peering in IX

[Adam Vitkovsky]

Hi,
Ok so you have already taken care of the TCP operation tuning so that the
MSS can be cranked up to max. 
Please be aware that the MSS is determined only during session initiation as
the TCP MSS option is carried only in the SYN and corresponding SYN/ACK. 
So you might need to bounce the session in order to get the advantage of the
new MSS determined by the PMTUD. 

BGP Update Generation. 
Should be taken care of automatically. 
Though it is good to check the number of update groups. 

Queue Optimization. 
1)
Selective Packet Discard (SPD) Headroom. 
Increase the SPD Headroom from default 1000. 
Increase the SPD Extended Headroom from default 10 -only if you see problems
with L2 keepalives. 

2)
(if the above does not help). 
Input hold queue. 
-increase the default value of 75 based on max possible number of TCP ACKs +
additional traffic for PRP. 
(if you still see input drops) . 
-increase min and max thresholds. 
-and turn on aggressive mode to only drop malformed packets. 

3)
System buffers. 
To handle a large influx of TCP ACK messages 64 bytes long thus stored in
small buffers. 
- Permanent buffers. 
- Min-free. 
- Max-free. 

I guess I'd have to refer you to the Cisco Press book: "BGP Design and
Implementation". 
Chapter 3; Tuning BGP Performance - in particular. 
Yes the book is old but lays on a great foundation on understanding all the
different aspects of BGP fine tuning. 


adam
-----Original Message-----
From: cisco-nsp [mailto:cisco-nsp-bounces at puck.nether.net] On Behalf Of
arulgobinath emmanuel
Sent: Wednesday, May 08, 2013 10:51 AM
To: Cisco Network Service Providers
Subject: [c-nsp] Any BGP fine tuning recommendation while Peering in IX

Hi all,
what are the common BGP fine tuning best practices  while peering more than
200 - 300 peering. except to the Path MTU , peer group i'm observing when
the RR flaps CPU goes high ( GSR 12406 / PRP-2/
12.0(33)S10  ) and due to that input queue on the interface goes high and it
causes random flap on almost half of the peer .

Thanks in advance,
Gobinath
_______________________________________________
cisco-nsp mailing list  cisco-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/cisco-nsp
archive at http://puck.nether.net/pipermail/cisco-nsp/