Slow througput

[Johnson Anthony]

We have a LAN based on 6509 core routers and the new one with 6807 core routers. The cross-segments traffic speed between the two cores is much slower, about 50% slower, than the traffic within the cores. The AutoQos and class-map commands seem to be different in both networks. Has anyone faced this issue? Can anyone throw some light on it? Thanks

Johnson Anthony

[Georg Pauwen]

Hello,

how do you have the link between the two cores configured ? Can you post a schematic drawing and indicate which traffic is slow ? Also, post the relevant configs...

[Johnson Anthony]

I have attached the simple version of the LAN:

The interlink router 6509 interface is:

interface TenGigabitEthernet2/8

description *** Migration Router Port to IT6807 TenGig1/32 ***

ip address 152.153.8.26 255.255.255.252

router eigrp 100

redistribute static

passive-interface default

no passive-interface TenGigabitEthernet2/8

network 152.153.0.0

network 172.25.0.0

no auto-summary

The interlink router 6807interface is:

interface TenGigabitEthernet1/32

description ***** Migration Router Port to CS6509 TenGig2/8 ****

ip address 152.153.8.25 255.255.255.252

router eigrp 100

network 152.153.0.0

network 172.25.0.0

redistribute static

passive-interface default

no passive-interface TenGigabitEthernet1/32

[Georg Pauwen]

Hello,

since you are using EIGRP over TenGigabitEthernet, consider using EIGRP Wide Metrics.

You will have to use name EIGRP instances. Your config would look like this:

6509(config)# router eigrp YourName
6509(config-router)# address-family ipv4 autonomous-system xxxx
6509(config-router-af)# network 152.153.8.24 0.0.0.3
6509(config-router-af)# metric weights 0 1 0 1 0 0 0

6807(config)# router eigrp YourName
6807(config-router)# address-family ipv4 autonomous-system xxxx
6807(config-router-af)# network 152.153.8.24 0.0.0.3
6807(config-router-af)# metric weights 0 1 0 1 0 0 0

Here is the reference document:

http://www.cisco.com/c/en/us/td/docs/ios-xml/ios/iproute_eigrp/configuration/15-sy/ire-15-sy-book/ire-wid-met.html

[Johnson Anthony]

Thanks. I will study the document and see if we I can implement these without disrupting the production network.

[Georg Pauwen]

Hello,

I guess you could use static routes while removing and reconfiguring the EIGRP process. Curious to know if this changes the throughput...

[Johnson Anthony]

Hi there,

Looks like I have got more than I have bargained for! The 3850 switches connected to the 6807 core themselves seem to have performance issues. I have connected two PCs to the ports on the same switch and the throughput is the same (50%) as the one across the cores. So I need to figure out if I am having some measurement issues or switch performance problems and resolve it before proceeding further.

Will keep you posted. Thanks.
 

[Johnson Anthony]

After all kinds of investigations, I have found several ports on the 3850 connected behind the 6807 core have Xmit-Err and coinciding with Output Errors/OutDiscards. So any PCs connected to these ports have very poor performances. I have changed QoS settings on these but to no avail. So the objective is to get to the bottom of the issue causing Xmt-Err & Output drops on the ports.

Any ideas??

MainAdminStack1#show interfaces counters errors

Port        Align-Err     FCS-Err    Xmit-Err           Rcv-Err   OutDiscards
Gi2/0/17            0           0           0           0          0            0
Gi2/0/18            0           0           0           0          0            0
Gi2/0/19            0           0           0           0          0            0
Gi2/0/20            0           0           0           0          0            0
Gi2/0/21            0           0           0           0          0            0
Gi2/0/22            0           0      115390           0          0       115390
Gi2/0/23            0           0      318454           0          0       318454
Gi2/0/24            0           0           0           0          0            0
Gi2/0/25            0           0           0           0          0            0
Gi2/0/26            0           0      348185           0          0       348185
Gi2/0/27            0           0           0           0          0            0
Gi2/0/28            0           0           0           0          0            0
Gi2/0/29            0           0     1623666           0          0      1623666
Gi2/0/30            0           0       74566           0          0        74566
Gi2/0/31            0           0           0           0          0            0
Gi2/0/32            0           0           0           0          0            0
Gi2/0/33            0           0           0           0          0            0
Gi2/0/34            0           0           0           0          0            0
Gi2/0/35            0           0           0           0          0            0
Gi2/0/36            0           0      693582           0          0       693582

[Johnson Anthony]

Looks like I will have to go with EIGRP Wide Metrics. If so, what happens to my original EIGRP configs? Does it remain same? and the new wide metrics applies only to the point to point 10 GBE interfaces? Does this mean I will have to also apply wide metrics to 40 GBE interfaces on the 40 GBE 6807 core? Your reply will be much appreciated. Thanks