cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
522
Views
0
Helpful
7
Replies

CEF per-packet load balancing performance issue

merryllem
Level 1
Level 1

Greetings All,

I'm scratching my head on a weird issue that I'm having. I have 2 OC3 point-to-point connection connecting site A and B. I have been transferring files from site A to B to see how much thruput is on my connection. Transfers from A to B is fine and dandy. The problem is when I transfer from B to A. The transfer from B to A is significantly (50%) slower than A to B (Same file and same servers). The weird part about this is when I disable one of my OC3's (disabled both)thruput seems to go UP. At this point I'm not quite sure where to start troubleshooting the problem.

Any insights would be appreciated

7 Replies 7

purohit_810
Level 5
Level 5

Are you Using Which routing protocol?

Can you observe performance by shutdown the one link completaly at both end?

Can check Duplex and speed on interfaces?

Errors?

If it is AUTO in mode duplex try to set hard set FULL bothe end?

Let me know...

Thanks,

Dharmesh Purohit

Are you Using Which routing protocol?

EIGRP

Can you observe performance by shutdown the one link completaly at both end?

when i shutdown one oc3 the perfomance(thruput) goes up (50%)

Can check Duplex and speed on interfaces?

Errors?

OC3

If it is AUTO in mode duplex try to set hard set FULL both end? n/a

Can u get the output of "sh ip cef detail"?

Rgds

site A

10.10.1.1/32, version 69, epoch 0, per-packet sharing

0 packets, 0 bytes

Flow: AS 0, mask 32

via 10.10.20.2, POS3/0, 0 dependencies

traffic share 1, current path

next hop 10.10.20.2, POS3/0

valid adjacency

via 10.10.30.2, POS3/1, 0 dependencies

traffic share 1

next hop 10.10.30.2, POS3/1

valid adjacency

0 packets, 0 bytes switched through the prefix

tmstats: external 0 packets, 0 bytes

internal 0 packets, 0 bytes

Site B

10.10.1.2/32, version 72, epoch 0, per-packet sharing

0 packets, 0 bytes

via 10.10.20.1, POS3/0, 0 dependencies

traffic share 1, current path

next hop 10.10.20.1, POS3/0

valid adjacency

via 10.10.30.1, POS3/1, 0 dependencies

traffic share 1

next hop 10.10.30.1, POS3/1

valid adjacency

0 packets, 0 bytes switched through the prefix

tmstats: external 0 packets, 0 bytes

internal 0 packets, 0 bytes

My endpoints (source and destination) would be a loopback IP's (10.10.1.2 (site A) & 10.10.1.1 (Site B)) on each router (l2tp)

a.alekseev
Level 7
Level 7

you use per-packet load balancing.

So this is you problem.

use per flow load balancing.

for link utilisation per-packet load balancing is good but it's bad for TCP

with per-packet load balancing you will have much higher CPU load. Do you check CPU load on both side during a file transfer?

this is the cpu usage when transfering the file

Site A

show processes cpu history table

CPU utilization for five seconds: 0%/0% at 00:00:09

CPU utilization for five seconds: 0%/0% at 00:00:04

Site B

show processes cpu history table

CPU utilization for five seconds: 1%/1% at 00:00:04

CPU utilization for five seconds: 0%/0% at 00:00:09

As far as checking speed duplex one thing you need to remember is that if you set one end to autonegotiate and the othet at 100 Full you will experience problems. The reson is that by default when setting up auto-negotiation it defualts to 100/Half duplex.HTH

Getting Started

Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: