[c-nsp] Couple performance questions regarding CAT 6500 (SUP 720-3BXL)

[Drew Weaver]

Hi list happy friday,

he BGP scanner issue has been beaten (literally) to death here, but I had a few general performance related questions regarding the 6500..

I notice that if I ping a somewhat busy interface on a 6500 about once a minute or so I get:

Reply from x.x.x.x: bytes=32 time<1ms TTL=253
Reply from x.x.x.x: bytes=32 time<1ms TTL=253
Reply from x.x.x.x: bytes=32 time=1ms TTL=253
Reply from x.x.x.x: bytes=32 time<1ms TTL=253
Reply from x.x.x.x: bytes=32 time=1ms TTL=253
Reply from x.x.x.x: bytes=32 time=172ms TTL=253
Reply from x.x.x.x: bytes=32 time=386ms TTL=253
Reply from x.x.x.x: bytes=32 time=366ms TTL=253
Reply from x.x.x.x: bytes=32 time=410ms TTL=253
Reply from x.x.x.x: bytes=32 time=353ms TTL=253
Reply from x.x.x.x: bytes=32 time=7ms TTL=253
Reply from x.x.x.x: bytes=32 time=66ms TTL=253
Reply from x.x.x.x: bytes=32 time=120ms TTL=253
Request timed out.
Reply from x.x.x.x: bytes=32 time=1ms TTL=253
Reply from x.x.x.x: bytes=32 time<1ms TTL=253

and it does seem to correspond to the BGP scanner running the CPU utilization up to 80%, is that the 'norm' for this time of high cpu utilization?

Second, 

I noticed we're having a high number of TTL failures:

TTL failures                          : 24541591

So I implemented a HW rate-limiter as such:

mls rate-limit all ttl-failure 500 10

Two questions about this,

A) is there any way to find out how many packets are being 'rate-limited' due to this command?
and
B) do I need to enable mls qos or anything else to 'globally enable' the HW rate-limiter?

3rd,

I'm noticing some queuing issues, 

Input queue: 0/75/13413/13085 (size/max/drops/flushes); Total output drops: 0
Output queue: 0/40 (size/max)
Input queue: 0/75/15112/15021 (size/max/drops/flushes); Total output drops: 0
Output queue: 0/40 (size/max)
Input queue: 0/2000/4294895378/0 (size/max/drops/flushes); Total output drops:4294941485
Output queue: 0/40 (size/max)
Input queue: 0/75/0/0 (size/max/drops/flushes); Total output drops: 0
Output queue: 0/40 (size/max)
Input queue: 0/2000/4294945720/0 (size/max/drops/flushes); Total output drops:1
Output queue: 0/40 (size/max)
Input queue: 0/2000/4294804008/0 (size/max/drops/flushes); Total output drops:3
Input queue: 0/75/549064/527178 (size/max/drops/flushes); Total output drops:2784
Output queue: 0/40 (size/max)
Input queue: 0/75/372439/361186 (size/max/drops/flushes); Total output drops:90049

I am using the Gig-E interfaces on the Sup720-3BXL as well as WS-X6724-SFPs

Is there a disadvantage to using the Interfaces on the SUP720-3BXL vs the 6724?

Should one modify settings to improve the queuing?

I was under the impression that the X6724 was not over-subscribed but from the looks of those queues it seems to be slightly inadequate.

any advice on any of these issues is greatly appreciated.

Thanks,
-Drew