>yup, we've done that one before, but not this time. Good idea, though. We
>had lots of ignoreds on the HSSIs prior to setting the mtu to 1500. I
>guess pool2 only has 85 4470byte buffers, where the pool1 1500 byte pool is
>much larger. These are wired pool sizes, I guess.
The buffers on the cbus are carved out at start-up time based on the MTU
allocations of all the interfaces on the box. These can not be changed
except by changing the MTU of interfaces and reloading.
If your buffer allocations on the cbus (MEMD) are not efficient, you can
get ignores, and in my experience, overruns... If you have several media
types that have different MTU's, you may need to adjust the MTUs of some
to get fewer pools with better allocation in MEMD.
Adjusting the hold queue on the interfaces should not have any affect as
far as I know, as those buffers are not the same as those in MEMD, correct?
If you either have a queuing strategy that can move MEMD buffers to
main memory (such as WFQ), or you turn it on (transmit-buffers backing-store)
then you may be able to buy yourself something. However, that may result
in slower switched packets, which may or may not be better than drops..
Sean
___________________________________
Sean Butler, CCIE #3897
IBM Global Services -- OpenNet Support
Phone: 8-631-9809, 813-523-7353
Fax: 8-427-5475 813-878-5475
Internet email: sebutler@us.ibm.com
This archive was generated by hypermail 2b29 : Sun Aug 04 2002 - 04:13:18 EDT