Please look at the simple schema I attached to this message.
I have lan that consist of a core 4503 and several 3750 access switches. They are all connected through the Gigabit ports, 802.1q trunk, VTP V1. They work in a cluster. Access switches have 24 * 100MBit/s ports and PCs are connected to them.
I have about 30 VLANs and their interfaces are configured as follows on the core switch:
VLAN10 - 10.0.10.1
VLAN20 - 10.0.20.1
VLAN30 - 10.0.30.1 etc.
I made a simple access map:
vlan access-map MAP1 10
match ip address ALL_TO_SERVER
vlan filter MAP1 vlan-list 2-1000
ip access-list extended ALL_TO_SERVER
permit ip any 10.0.10.0 0.0.0.255
permit ip 10.0.10.0 0.0.0.255 any
permit ip 10.0.20.0 0.0.0.255 any
permit ip any 10.0.20.0 0.0.0.255
Clearly I want VLAN10 & 20 to be accessible by everyone and vice-versa.
The problem involves speed.
When I set a PC to a VLAN10 on the second(right) 3750 all works fine - 100MBit/s
But when I place it in VLAN20 or any other VLAN different than VLAN10 the speed falls to 16Mbit/s.
Since I have limited number of users i do not worry of network congestion.
I just want to disable all kind of QoS or shaping.
I tried :
no qos - on 4503
no mls qos - on 3750
I also tried disabling features as DBL, header rewrite etc. I have nothing specified per switch port that might apply some kind of bandwidth restriction.
Unfortunatelly nothing works.
May be it is some setting involving intra-VLAN communication?
Any help or guideline will be appreciated. Thanks.
you are facing a clear performance issue in intervlan routing.
However, this shouldn't be related with QoS enabled or not.
Verify the path of routed packets, default gateways on the PCs.
Also the sh proc cpu sorted 1 min on the C4503 can help.
You may like to post sh outputs and selected parts of configuration.
Also it is important to know how you are performing this performance tests?
are using FTP sessions, TCP negotiated sessions with netperf or ipperf?
Hope to help