Wireless performance between different VLANs

[tiago.molinos]

Hi,

I have a problem regarding throughput over wireless when the peers of the data transfer are in different VLAN (separate networks).

Topology:

I have a 3560 L3 switch configured with two VLANs, VLAN1 is the PC Wired network and VLAN2 is the PC Wireless network.

Connected to this switch I have a 1602i standalone access-point.

I have connected two PCs via Gigabit Ethernet ports to both VLANs (wireless is disabled).

I have one wireless client connected via 802.11an (disabled 802.11bgn to rule out interferences) using 40MHz channel width in the wireless network (VLAN2)

Device description:

PC1 - Connected via Gb ethernet to VLAN1;

PC2 - Connected via Gb ethernet to VLAN2;

PC3 - Connected via Gb ethernet to VLAN1;

WPC1 - Connected via 802.11n to VLAN2;

Here's the the tests I've made:

I've used both iperf and multiple paralel FTP sessions with similar results.

PC1 -> PC3

447 Mbits/sec

Note: no routing involved, no wireless involved

PC1 -> PC2

220 Mbits/sec

Note: inter-vlan routing involved, no wireless involved

PC2 -> WPC1

80 Mbits/sec

Note: no routing involved

PC1 -> WPC1

9.8 Mbits/sec <== PROBLEM

Note: inter-vlan routing involved

As you can see I have nearly a 10x factor decrease in performace in the last test presented. I can't find why this happens!

I've already updated both the access-point and the switch to the latest version avaliable.

I've also used an old AP1100G and got the same result, around 9.8Mbits/sec when in different networks and around 17Mbit/sec when in the same network.

I've tested with diffent security settings in the wireless ssid, but I always get the same results... a huge performance degradation when connecting two devices in different VLANs.

Any suggestion? Thanks!

[Scott Fella]

Well you ruled out the wireless somewhat because you still get poor performance on the wireless to wired with a different AP. Is that switch also doing the routing or is other equipment involved? That is what I would look at because Vlan to Vlan on wired should be the same.

Sent from Cisco Technical Support iPhone App

[tiago.molinos]

Thank you for your answer. I wanted to keep the lab as small as possible, so the routing is done by the 3560. I can see that while routing between two different VLANs the device is at least capable of doing 220Mbits/sec so I think this might be somehow a limitation on the routing capacity in this switch. Still 220Mbits/sec is way more than the nearly 10Mbits/sec i'm getting...

[Scott Fella]

Here is the thing. Update the driver for the wireless card on the PC.

Test on the 2.4ghx only
Test on the 5ghz with 20mhz channel width

See if any of these make a difference.

Sent from Cisco Technical Support iPhone App

[tiago.molinos]

Ok,

I shut down the 5GHz Interface, and enabled the 2,4GHz interface. Got the same result (8.46Mbit/sec).

Got back to the 5GHz interface using only 20Mhz width, got (9.37 Mbit/sec).

It took me a bit more of time updating the driver. I thought maybe it could be it. I was using Microsoft Driver.

Changed it to Intel drivers, but got the same result (9.38 Mbits/sec). I'll try to test with different hardware I get my hands in a different notebook.

So... none of your suggestions mada a difference.

[Scott Fella]

You are just eliminating the wireless is what your trying to do.

Sent from Cisco Technical Support iPhone App

[Rasika Nayanajith]

Hi Tiago,

First of all I would ensure client is connecting with correct bandwidth. Below command output should verify client is assoicated in 5GHz & 40MHz channel.

AP#show dot11 associations

Then I would do the test with upstream & downstream direction seperately. ie PC1 ->WPC1 & WPC1-> PC1. Would like to see what you get. Do the same for PC2-> WPC1 & WPC1-> PC2 to compare.

Hope you got only this WPC1 client associated to your AP while you are doing this test. Confirm that as well.

HTH

Rasikka

**** Pls rate all useful responses ****

[tiago.molinos]

Hi!

Thank you for your answer,

as you pointed out, there are very large differences dependant of the stream's direccion.

Here are the results:

PC1 -> WPC1

9.8 Mbits/sec

WPC1 -> PC1

76.52 Mbits/sec

PC2 -> WPC1

56 Mbits/sec

WPC1 -> PC2

85.79 Mbits/sec

There are no other wireless clients involved, but it's obviously direction dependant...

[Scott Fella]

Well those are interesting results... well we know that its not the AP because you see the same results using an AP1100, but from wired to wired it seems to be 50% of the throughput when your crossing vlans, Maybe test with two wireless clients on the same vlan and then crossing vlans and see if you see the same 50% degrade from two wireless machines.

Thanks,

Scott

Help out other by using the rating system and marking answered questions as "Answered"

[Scott Fella]

Your switch is pretty much default configs or are you doing any ACL's or policing?

Thanks,

Scott

Help out other by using the rating system and marking answered questions as "Answered"

[Rasika Nayanajith]

Hi Tiago,

Resutls proves that you have issue only in downstream direction to the wireless client when traffic originating from a different subnet to wireless client.

Do you have any sort of QoS configured on your 3560 switch ? (show mls qos should verify it ) I would like to see the "show run  interface gx/x" of the AP connected switchport & show run int vlan 1 & show run int vlan 2 output as well.

HTH

Rasika

*** Pls rate all useful responses ****

[tiago.molinos]

Hi! Thank you for you answers,

Here is the output of the commands:

Switch#sh mls qos

QoS is disabled

QoS ip packet dscp rewrite is enabled

Switch#sh run int g0/13

Building configuration...

Current configuration : 150 bytes

!

interface GigabitEthernet0/13

switchport trunk encapsulation dot1q

switchport trunk native vlan 2

switchport mode trunk

load-interval 30

end

Switch#sh run int vl 1

Building configuration...

Current configuration : 89 bytes

!

interface Vlan1

ip address 10.10.10.125 255.255.255.128

end

Switch#sh run int vl 351

Building configuration...

Current configuration : 117 bytes

!

interface Vlan351

description Wireless LAN

ip address 10.10.10.158 255.255.255.224

end

Next I'll try to conduct the test between two wireless clients.

[Rasika Nayanajith]

Thanks for this output, it confrims there is no issue with the switch end configuration.

Yes do the test between two wireless clients & see whether problem exist downstream direction towards wireless client.

Other thing I would suggest it to check with lateste IOS image & see.

HTH

Rasika

**** Pls rate all useful responses *****

[tiago.molinos]

Both PCs on Wireless

Same VLAN

WPC1 (VLAN2) -> WPC2 (VLAN2)

39.19 Mbit/sec

WPC2 (VLAN2) -> WPC1 (VLAN2)

31.31 Mbit/sec

Different VLANs

WPC1 (VLAN1) -> WPC2 (VLAN2)

9.15 Mbit/sec

WPC2 (VLAN2) -> WPC1 (VLAN1)

37.36 Mbit/sec

Both PCs on Wired

Same VLAN

PC1 -> PC2

670 Mbit/sec

PC2 -> PC1

705 Mbit/sec

Different VLAN

PC1(VLAN2) -> PC2

214 Mbit/sec

PC2 -> PC1 (VLAN2)

9.8 Mbit/sec !!! (Here's the real problem!!!)

So I guess I can rule out this beeing a wireless issue! Still have to find the problem anyway... I missed this result the first time because I was only testing one direction...

I've allready upgraded the switch to the latest IOS release, as I have with the access-point...

[Scott Fella]

You can baseline the average when the wired devices are on the same Vlan and then look at both results when you are crossing a layer 3 boundary. Your throughput drops a lot in both scenarios which it shouldn't. The one that has 9mbps is even worse, but the other one is still low.

Sent from Cisco Technical Support iPhone App