Cisco 2651XM high cpu usage 12.4(25b)IPBASEK9

Hello I have a problem:

I have one C2651XM 48F/256D with one WIC-ADSL on a 7M/512k ADSL LINE

Just a simple ACL inbound, NAT1-1 towards client and nothing else. I get at 3mbit throughput download, about 25% CPU usage.

It seem very high to me, since I plan to put later a second WIC ADSL to use a second adsl line.

(actually in the router there are 4 wic adsl, but two will be removed , and the NM-2W with them).

IOS is 12.4(25)b IPBASEK9

This is my current conf:

gw1.wisp#sh run Building configuration...

Current configuration : 7834 bytes ! version 12.4 service nagle no service pad service tcp-keepalives-in service tcp-keepalives-out service timestamps debug datetime msec localtime show-timezone service timestamps log datetime msec localtime show-timezone service password-encryption service internal service sequence-numbers ! hostname gw1.wisp ! boot-start-marker boot-end-marker ! logging buffered 65535 notifications no logging console no logging monitor enable password 7 xxxxxxxxxxxxxx! no aaa new-model clock timezone CET 1 no network-clock-participate slot 1 no network-clock-participate wic 0 no ip source-route no ip gratuitous-arps ip cef ! ! ! ! no ip bootp server ip domain round-robin ip domain name spadhausen.com ip name-server 212.97.32.2 ip name-server 94.141.24.92 login block-for 120 attempts 5 within 60 login on-failure log ! ! ! username zzz privilege 15 password 7 xxxxx archive log config hidekeys ! ! ip tcp ecn ip tcp selective-ack ip tcp timestamp ip tcp window-size 65535 ip tcp synwait-time 10 ip ssh time-out 90 ip ssh version 2 ! ! interface Null0 no ip unreachables ! interface ATM0/0 description xxx no ip address no ip redirects no ip proxy-arp atm restart timer 300 atm ilmi-keepalive dsl operating-mode auto clock rate aal5 7000000 ! interface ATM0/0.1 point-to-point ip address 89.186.xxxx 255.255.255.252 ip access-group 100 in no ip unreachables no ip proxy-arp ip nat outside pvc 8/35 encapsulation aal5snap ! ! interface FastEthernet0/0 description Interfaccia verso la backbone interna ip address 172.16.0.1 255.255.254.0 no ip redirects no ip proxy-arp ip nat inside duplex auto speed auto ! interface ATM0/1 no ip address shutdown atm restart timer 300 no atm ilmi-keepalive dsl operating-mode auto clock rate aal5 7000000 ! interface FastEthernet0/1 no ip address shutdown duplex auto speed auto ! interface ATM1/0 no ip address shutdown atm restart timer 300 no atm ilmi-keepalive dsl operating-mode auto ! interface ATM1/1 no ip address shutdown atm restart timer 300 no atm ilmi-keepalive dsl operating-mode auto ! interface Dialer0 no ip address no cdp enable ! ip forward-protocol nd ip route 0.0.0.0 0.0.0.0 ATM0/0.1 ip route 172.16.2.0 255.255.255.0 172.16.0.254 ip route 172.16.3.0 255.255.255.0 172.16.0.254 ip route 172.16.4.0 255.255.255.0 172.16.0.254 ip route 172.16.5.0 255.255.255.0 172.16.0.254 ip route 172.16.6.0 255.255.255.0 172.16.0.254 ip route 172.16.7.0 255.255.255.0 172.16.0.254 ip dns server ! no ip http server no ip http secure-server ip http timeout-policy idle 60 life 86400 requests 10000 ip nat translation timeout 3600 ip nat translation tcp-timeout 1200 ip nat translation udp-timeout 100 ip nat translation finrst-timeout 15 ip nat translation syn-timeout 45 ip nat translation icmp-timeout 120 ip nat inside source list 101 interface ATM0/0.1 overload here just some nat 1-1 ! ! no logging trap access-list 99 permit 77.93.235.238 access-list 99 permit 172.16.0.0 0.0.1.255 access-list 101 permit ip 172.16.0.0 0.0.1.255 any snmp-server community public RO 99 snmp-server ifindex persist snmp-server location MG Casa snmp-server contact snipped-for-privacy@spadhausen.com no cdp run ! control-plane ! banner login ^C You are connected to $(hostname).$(domain) on line $(line). If you are not authorized to access this system, disconnect now.

THIS IS FOR AUTHORIZED USE ONLY

Unauthorized or improper use of this system may result in administrative disciplinary action and civil and criminal penalties. By continuing to use this system you indicate your awareness of and consent to these terms and conditions of use. LOG OFF IMMEDIATELY if you do not agree to the conditions stated in this warning.

Network Administrator: snipped-for-privacy@spadhausen.com ^C ! line con 0 login local line aux 0 line vty 0 4 privilege level 15 login local transport input ssh ! scheduler max-task-time 5000 ntp clock-period 17208075 ntp server 192.43.244.18 ntp server 193.204.114.105 ! end

Reply to
Elia S.
Loading thread data ...

It doesn't seem that high of CPU usage to me for a 26xx class router. There's not exactly alot of horsepower in there. It was designed to drive maybe 3 T1s.

Reply to
Doug McIntyre

Hello I have noticed that from 3mbit traffic to 4mbit the cpu usage raises at about 28-30%, so it is not a regular increase.

Could I expect 14mbit full throughput, when downloading from 2x WIC ADSL with 7M/512K line without saturating the cpu ?

"Doug Mc>I have a problem:

It doesn't seem that high of CPU usage to me for a 26xx class router. There's not exactly alot of horsepower in there. It was designed to drive maybe 3 T1s.

Reply to
Elia S.

Meanwhile, at the comp.dcom.sys.cisco Job Justification Hearings, Elia S. chose the tried and tested strategy of:

If you mean that 4Mbps requires 30% CPU, assuming a linear relationship you'd still need 105% CPU to get 14Mbps. If it isn't linear [which you've indicated it may not be] then who knows.

How about testing the CPU usage NATing from one ethernet to another? If this requires less CPU you could stick a basic DSL router on one of the ethernets instead of using a DSL WIC.

Reply to
alexd

I wouldn't expect over 8Mbps-10Mbps with NAT on any 26xx class router..

Reply to
Doug McIntyre

I confess that I have never done NAT vs noNAT performance testing however I have never thought it much of a hit.

Hmmm. Here is what cisco say about 2651XM.

Cisco Router performance

formatting link
to work without login? If it needs login search site/google for document. This is a VERY handy document and often almost impossible to find by searching for keywords. Worth making a note of.

Process Switching _ Fast/CEF Switching PPS . Mbps _ PPS . Mbps

265x(XM) 2,000 1.024 40,000 20.48

- Mbps calculated by pps * 64bytes * 8bits/byte

In software routers (not switches like Catalyst 3750) forwarding is limited by packet rate, packet size in effect does not matter.

So 2651XM can do at least 20Mbps. With an average packet size of

200 bytes it will be about 60Mbps.

Please put router under load and post

sh proc cpu sh interfaces switching

May as well post sh int for DSL and Ethernet too.

you CPU is

25/3 =3D 8 percent per Mbps 28/4 =3D 7 percent per Mbps 30/4 =3D 7.5 percent per Mbps

At 100% we might expect 12Mbps

All reasonably consistent although CPU seems higher than cisco are predicting. Oh yes! Maybe a lot of packets are being fragmented?

sh ip traffic - please post this too.

There is nothing wrong with fairly high CPU. Since you are not doing any dynamic routing I would not worry at all at 70%. That is what the cycles are for:)

You havent left any debugs running?

Reply to
bod43

the generic page shows some related performance info. I dont need to log in to get to it either.

formatting link

The 20 Mbps is for 64 byte packet size and "raw" IP routing only.

Turning on features like NAT will have some impact, and if you have any QoS or shaping that will degrade the expected throughput.

A software based router will show some fall off in performance as packet size increases, but as said it should be relatively minor.

Performance numbers are usually measured on high speed interfaces such as Ethernet - others can "cost" more to drive, so ADSL may reduce throughput

I think there is a performance limit between a WIC card and the box, so the ADSL interface may limit throughput.

You talk about using dual ADSL WICs - that is going to need some sort of load balancing - again likely to degrade the achievable performance.

Reply to
Stephen

Hello I will do all the test suggested and will post the info when I will pe hand on the router.

I will answer some questions below:

"Stephen" ha scritto nel messaggio news: snipped-for-privacy@4ax.com...

I tried removing NAT and routing all the traffic to another router, and I gained a lot of CPU load, the issue should be the WIC slot's BUS that is driving a lot of CPU.

I saturated cpu of the 2651xm routing between the two fast eth interfaces and doing inspection (match protocol) and it saturated at about 45mbit!!! But was FE to FE.

No load balancing, but two separate ADSL links with separate /28 ip addresses on them , and PBR (later) to decide who to go out where.

I did not undestand that.

Reply to
Elia S.

It's a "plot" of your 'CPU load' vs 'throughput'.

You said 25% at 3Mbps You said 28% at 4Mbps OR

30 at 4Mbps

For each of your samples I computed the percent CPU per Mbps.

So for the very limited sample it seems that we have

7 or 8 percent CPU per Mbps throughput.

As I said post sh proc CPU for router under load. there is a lot in there that can show what is going on.

Reply to
bod43

the critical bit of a cisco IOS router is keeping packet throughput using the high efficiency processing for any high load traffic.

If you want PBR to handle a significant load, (and it will only be outbound traffic, so limited by the ADSL uplink speed), then you need an IOS that can CEF switch PBR traffic.

you will probably need any ACLs that choose what goes via PBR to be fairly simple as well.

If you dont manage this, then the PBR traffic can still go out the interface, but using the "process" switching path - maybe 100x slower than CEF, but more importantly creating a big hit on the CPU load.

Reply to
Stephen

Cabling-Design.com Forums website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.