10-30-2010 09:52 AM - edited 03-06-2019 01:48 PM
We noted a very low rate of transfer in the stack when the stack has to do inter vlan routing, if the PCs are in the same vlan everything works fine.
The stack does not have a complicated setup, QoS is not applied to the ports with which we tested, and access lists, nothing more
I have read online and there are more people with this problem. Anyone can help me?
Switch Ports Model SW Version SW Image
------ ----- ----- ---------- ----------
* 1 12 WS-C3750G-12S 12.2(35)SE5 C3750-IPBASE-M
2 12 WS-C3750G-12S 12.2(35)SE5 C3750-IPBASE-M
3 28 WS-C3750G-24TS-1U 12.2(35)SE5 C3750-IPBASE-M
4 28 WS-C3750G-24TS-1U 12.2(35)SE5 C3750-IPBASE-M
Thanks in advance.
10-30-2010 10:45 AM
Hello Oscar,
Interesting. Does this happen for all inter-VLAN routing, or is only a particular combination of VLANs (from-to) affected?
For a multilayer switch in elementary configuration, intra-VLAN switching shall be as effective as inter-VLAN routing. The reason for a dramatic decline in the throughput of the routed traffic is often caused by the fact that the traffic has to be processed by the CPU, instead of being routed in hardware. IP packets that cannot be routed in hardware include packets requiring fragmentation (check the MTU settings!), IP options, IP packets requiring ICMP responses and other packets that are unsupported within the CEF/TCAM infrastructure.
Try using the show cef not-cef-switched and show ip cef switching statistics commands to verify the counts of punted packets (packets sent to the CPU instead of being switched in hardware). High counts of these packets increasing intensively in time indeed suggest that the traffic is sent to the CPU. Also verify if there are any punt-type adjacencies installed in the TCAM using the show ip cef adjacency punt command - ideally, there shall be none.
Also, what is the SDM template you are using? Is there enough space in the TCAM for your IPv4 routing information? Verify it using the show platform tcam utilization to see if the "Used Masks/values" column reports a smaller usage for IPv4 routes than the "Max Masks/Values" column. If a route can not be installed into the TCAM, the packets will again be punted.
Best regards,
Peter
10-30-2010 11:01 AM
Thanks for your answer Peter
We have several VLANs, but they all have only communication with VLAN2
VLAN2 interface
ip address 10.43.20.49 255.255.255.0 secondary
ip address 10.43.22.49 255.255.255.0 secondary
ip address 10.43.26.49 255.255.255.0 secondary
ip address 10.43.21.53 255.255.255.0
no ip unreachable
If I connect a PC to gi 4/0/23 vlan 2 with direction of range 10.43.20.X and other PC to gi 4/0/24 vlan 2 with IP of range 10.43.21.X, transfer rate very very slow
If I connect a PC to gi 4/0/23 vlan 2 with direction of range 10.43.22.X and other PC to gi 4/0/24 vlan 2 with IP of range 10.43.26.X, transfer rate its good
(without level 3) --> If I connect a PC to gi 4/0/23 vlan 2 with direction of range 10.43.21.X and other PC to gi 4/0/24 vlan 2 with IP of range 10.43.21.X, transfer rate its good
If I connect a PC to gi 4/0/23 change to vlan 3 with direction of range X.X.X.X and other PC to gi 4/0/24 vlan 2 with IP of range 10.43.21.X, transfer rate very very slow
The tests tell me I'm not going to be able to do until Tuesday.
Best Regards.
10-30-2010 11:04 AM
Hello Oscar,
Thank you for the information you've provided. I will have to go over it more carefully. In the meantime, please, can you verify the several show commands I have suggested in my first post? These are not disruptive and can be performed anytime.
EDIT: I apologize, I did not understand you won't be able to perform those show commands until Tuesday. I apologize - no problem, we'll wait till then.
Best regards,
Peter
11-02-2010 03:06 AM
Hello!!
I managed to reproduce the problem in laboratory with 1 only swicht
The transfer rates are very low with this test (attached transfer rate)
PC1 10.43.21.50
DG 10.43.21.53----- Gb 5/0/23 (vlan2) SWITCH ----Gb 5/0/23 (vlan2) PC2 10.43.20.50
DG 10.43.20.49
Backbone#show run int vlan 2
Building configuration...
Current configuration : 256 bytes
!
interface Vlan2
ip address 10.43.20.49 255.255.255.0 secondary
ip address 10.43.22.49 255.255.255.0 secondary
ip address 10.43.26.49 255.255.255.0 secondary
ip address 10.43.21.53 255.255.255.0
no ip unreachables
11-02-2010 04:06 AM
HelloT
he tests I've done in the LAB are not reliable, I have 2 PCs directly connected I have the same transfer rate, I'm trying to send me the output of the commands you asked me.
Regards.
11-02-2010 04:58 AM
Hello
These are the commands on the switch that has the problem:
show cef not-cef-switched
% Command accepted but obsolete, see 'show (ip|ipv6) cef switching statistics [feature]'
IPv4 CEF Packets passed on to next switching layer
Slot No_adj No_encap Unsupp'ted Redirect Receive Options Access Frag
RP 546463 0 546483 0 0 0 0 0
3 0 0 0 0 0 0 0 0
2 0 0 0 0 0 0 0 0
1 0 0 0 0 0 0 0 0
Backbone#
Backbone#show ip cef switching statistics
Reason Drop Punt Punt2Host
RP LES No route 6 0 0
RP LES No adjacency 87576 0 546488
RP LES Incomplete adjacency 0 0 1
RP LES TTL expired 0 0 19
RP LES Total 87582 0 546508
All Total 87582 0 546508
Backbone#
Backbone#show ip cef adjacency punt
Prefix Next Hop Interface
Backbone#
Backbone#show platform tcam utilization
CAM Utilization for ASIC# 0 Max Used
Masks/Values Masks/values
Unicast mac addresses: 784/6272 79/571
IPv4 IGMP groups + multicast routes: 144/1152 6/26
IPv4 unicast directly-connected routes: 784/6272 79/571
IPv4 unicast indirectly-connected routes: 272/2176 17/104
IPv4 policy based routing aces: 0/0 0/0
IPv4 qos aces: 528/528 18/18
IPv4 security aces: 1024/1024 110/110
Note: Allocation of TCAM entries per feature uses
a complex algorithm. The above information is meant
to provide an abstract view of the current TCAM utilization
Backbone#
Regards.
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