12-02-2008 09:07 AM - edited 03-06-2019 02:46 AM
Hello,
below it output from Sup720 on 6513 Catalist (native).
Does anybody know what the first line stand for?
"Port manager per"
it shows some numbers when CPU goes above 50%,
shows almost nothing when CPU is below 20%
sh proc cpu sorted
CPU utilization for five seconds: 58%/50%; one minute: 53%; five minutes: 53%
PID Runtime(ms) Invoked uSecs 5Sec 1Min 5Min TTY Process
303 8278728 50833345 162 3.11% 1.44% 1.37% 0 Port manager per
146 8210348 39634349 207 1.43% 1.05% 1.15% 0 IP Input
20 2980640 27745787 107 0.95% 0.25% 0.23% 0 IPC Seat Manager
346 4080588 9354823 436 0.71% 0.59% 0.60% 0 SNMP ENGINE
344 2682360 11436871 234 0.47% 0.27% 0.30% 0 IP SNMP
8 677512 2476194 273 0.47% 0.14% 0.12% 0 ARP Input
199 292732 660609 443 0.39% 0.24% 0.22% 0 CEF: IPv4 proces
110 80996 124441 650 0.07% 0.04% 0.05% 0 HC Counter Timer
52 219280 11625169 18 0.07% 0.01% 0.00% 0 Net Input
345 522052 5406766 96 0.07% 0.05% 0.05% 0 PDU DISPATCHER
10 0 2 0 0.00% 0.00% 0.00% 0 AAA high-capacit
9 0 1 0 0.00% 0.00% 0.00% 0 AAA_SERVER_DEADT
11 0 1 0 0.00% 0.00% 0.00% 0 Policy Manager
7 0 2 0 0.00% 0.00% 0.00% 0 Timers
15 0 6923 0 0.00% 0.00% 0.00% 0 IPC Dynamic Cach
And another question, why numbers differ so much,
summarizing from the list I can hardly get 5 - 10%,
why summary line shows 50 and above ?
Thank you
Alex
12-02-2008 12:19 PM
I don't know the purpose of "Port manager per" process, to answer your first question, but for your second: the first two numbers, 58%/50%, normally represent total CPU and interrupt CPU. The second, interrupt, should be fast path software forwarding of packets. The difference, in this case 8%, would be all other software processes. I.e. the total of all the process CPU percentages should be about 8%.
12-02-2008 12:33 PM
Joseph,
Is this Fast Path software forwarding related to
physical ports on blades forwarding traffic?
Thanks
Alex
12-02-2008 12:54 PM
On software based routers, it is and it's the ideal. With your 6513, I'm unsure since much of normal forwarding is done within hardware. On this system, it might indicate forwarding that couldn't be done by the hardware and had to be done by software, although in an optimal switching path, or it might indicate a pretty busy sup720. Perhaps someone else will respond on high "interrupt" CPU meaning on the 6500 (or 7600) platform.
12-02-2008 12:57 PM
That performance could be improved by putting DFC on the modules, could I say that?
Right now there are no DFC's in there.
Thanks
Alex
12-02-2008 01:04 PM
It might if the issue is indeed the sup720 has a high load. (Not real hard to do today with high density gig line cards.)
Also if your line cards support DFCs, not all do. The best would be the CEF720 cards, since they would have the most bandwidth to the sup720 fabric, but I believe except for the 24 port SFP card (6724-SFP), they require dual fabric connections which your 6513 only supports in slots 9-13.
12-02-2008 01:09 PM
This is what we have, dual 720, NAM, all top (2 - 6) are 24-SFPs, all bottom (9 - 13) are 48-Gig RJ-45, dual attached ones. All except NAM can bear DFC's.
Thanks
Alex
12-02-2008 01:58 PM
In that case, DFCs in one or more cards might help. If most of the load was on just one or a couple of cards, those would likely obtain the most benefit with DFC on their cards. I.e., might not be real need or benefit to upgrade all cards.
However, would still like to see someone more familar with 6500 series comment. Also, I think other commands (sh controller?) might indicate further where (what cards or ports) your 6513 might be busy.
12-02-2008 02:01 PM
Thank you Joseph,
I appreciate your input
Alex
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide