Improving Wireless Performance

[Stephen Carville]

I have been getting lot of complaints from the C-level about the speed of the wireless network.  Tests on the wired network yield 700+ Mbps but the wireless strains to reach 24 Mbps.

I do not expect the wireless network to match the wired network in speed or reliability but the above does seem a bit of an excessive difference.  Is there any good documentation on improving wireless performance?

Controller: 5508 (two configured as active-passive)
Access Points: AIR-CAP3702E (17 local; 29 total)
All APs are in Flex-Connect mode.

[Stephen Carville]

---

@patoberli wrote:

Ok great. What is the channel width you have configured on the 5 GHz interface of your test AP? Please note 30 and 43 Mbit/s are fairly close to the possible maximum you get with an 802.11n 1SS client on 2.4 or 5 GHz with a 20 MHz channel. That is in a somewhat busy environment. In a basement you can reach up to around 70 Mbit/s with the same test client.

Right now it is set to 40 MHz. I also tried it at 80 MHz but I guess neither of my laptops will connect to mutiple channels. They are both castoffs nobody wanted any more so maybe they are just too old. Next week I will try to get a Windows 10 laptop to test from.

The world looks way different if you go for 80 MHz channels and a 3SS client which runs 802.11ac, that can reach up to 500 Mbit/s realistically.

If I could get the bigshot laptops to operate at 500 Mb/s I'd be the corporate hero of the day. :)

You shouldn't use encrypted protocols (like SFTP) to test bandwidth, because the encryption can be very heavy on the client CPU and produce wrong data.

I've found sftp between Linux boxes to be not too much slower than FTP or NFS.

[patoberli]

Ok, keep us updated about the results with a current device.
It ideally has an 802.11ac Wave2

[Stephen Carville]

Helpdesk has not yet brought me the promised Windows laptop but I was able to grab a Mac and test with it.  Using iperf I was able to reach 200 Mb/2 to 400 Mb/s.  That's with 80 MHz bandwidth.  Not as good as I'd like it but way better than before.

[patoberli · Accepted solution]

Now you're slowly getting to the possible maximum. Much more will not be possible with your test hardware and neither can the AP offer much more (in a just slightly busy wireless environment).
In any case, here is the Cisco Tuning Guide for High Density networks. It contains a lot of valuable information:
https://www.cisco.com/c/en/us/products/collateral/wireless/aironet-1250-series/design_guide_c07-693245.html
It's not fully up to date (lacking 802.11ac), but most that was valid for the older standards is still valid for AC. Only that you need 80 MHz for the really high throughput and capable clients is new.

[Stephen Carville]

In the testing environment, both Mac and Windows 10 laptops were able to reach 200 Mbps the 400 Mbps.  Speeds are a little less in the real world but still much better than before. 

Thanks to everyone who helped.

[Max Jobs]

Thank you Pato, can you share the newest version of this guide?

[patoberli]

I don't work for Cisco, but maybe those guides here help:
https://www.cisco.com/c/en/us/products/collateral/wireless/catalyst-9800-series-wireless-controllers/guide-c07-743627.html
https://www.cisco.com/c/en/us/td/docs/wireless/controller/technotes/8-7/b_wireless_high_client_density_design_guide.html