Hi Kanako,
understand your question, but something like that doesnt exist for the trident/typhoon NPU's for the asr9k.
in fact cpu usage is not really applicable for parallel/serialized hardware forwarders, but mainly the KPI is here
the PPS it handles.
The max pps it can handle is dependent on feature set.
What you could do is measure the rates on the PARSE_FABRIC_RECEIVE count (packets received from fabric, so egress direction), PARSE_ENET_RECEIVE count (packets received from the interface to the fabric, so ingress) and the LOOPBACK count, that is packets that are recirculated for extra processing. The sum of these 3 gives an idea of the PPS rate the NPU is working on.
Memory consumption is something tricky also, there are many memories attached to the NPU, notably:
TCAM used for vlan matching, ACL and QOS matching
Search memory for L2 and L3 (shared on Trident, separated on Typhoon)
Frame memory for packet bufferin in QOS
Stats memory, for all kinds of counters of QOS maps, interface stats etc
their usage is rather dynamic also.
The few things you want to verify on these are:
- ACL usage (in terms of Tcam entries)
- MAC table size (2M max on typhoon)
- FIB size (4M max on typhoon)
frame memory is dependent on the buffering, which is a sum of all the packets instant queue length and size of all pmaps attached to interfaces served by the same NPU. there is no good overview as to how much you have available of that, but it has a rather significant size 1G or 2G depending on the TR or SE cards (typhoon specific).
regards
xander
Xander Thuijs CCIE #6775
Principal Engineer
ASR9000, CRS, NCS6000 & IOS-XR
