Slow iPerf tests and WiFi speeds on AP230

  • 2
  • Question
  • Updated 2 years ago
  • (Edited)
I work at a school with a mix of AP230, AP121 and AP120 APs and we have been receiving slow WiFi performance. At first, I was trying to look at interference, channels, basic data rates, DAS etc but I then found out that the APs have a built in iPerf 2 client/server.

The AP I am testing with is a AP230 with 2 aggregated links to a gigabit switch and a desktop PC on the same gigabit switch. I run the command "iperf server" on the AP via SSH. I can achieve over 900 Mbits/s to other hosts on the network from this PC, but when I test to the AP, I get a maximum of 286 Mbits/s. The CPU and RAM usage is staying low. I can make loads of WiFi optimisations, but it looks as if the WiFi is being bottlenecked by the LAN performance.

After getting these results, I manually set both ports to 1000 full duplex via HiveManager which did not make a difference. We love the features of HiveManager and the control it gives you, but the speeds are not up to par.

We are running HiveOS 6.4r1d.2111 on the APs. Can anyone try testing their APs and seeing if this is re-producible? Is there a switch that needs flicking somewhere to increase the performance?

EDIT : Has anyone experienced a performance improvement switching to HiveManager NG?
Photo of Mr B Williams

Mr B Williams

  • 3 Posts
  • 0 Reply Likes

Posted 2 years ago

  • 2
Photo of Tony Schaps

Tony Schaps

  • 28 Posts
  • 8 Reply Likes
I am intrigued to try this on my network, but in the meantime, did you try this without the aggregated links? In my non-Aerohive experience, link aggregation adds a layer of potential confusion to packet switching algorithms if not perfectly implemented by the manufacturer, and it would not take much of a delay to have the effect you're seeing. I would try it first by just physically unplugging one of the links, but it would be best to try it with a single link and aggregation config removed. If you already have tried that, please confirm. 
Photo of Mr B Williams

Mr B Williams

  • 3 Posts
  • 0 Reply Likes
I can confirm that I have also tried this without link aggregation with the config removed. I would also like to point out that this is happening on all of our APs including AP230, AP121 and AP120, most of which do not have link aggregation.

Thanks very much for trying this also. Once you give us your results, can you also confirm the firmware version on your APs to see if they match?

Thanks,
Brandon
Photo of Tony Schaps

Tony Schaps

  • 28 Posts
  • 8 Reply Likes
Sorry, I understood this was only the 230s. I probably will not have time to try it until later in the week, but I can tell you that I recently updated my APs from 6.4 to 6.5 (HiveOS 6.5r3a Honolulu.2530 on the AP230) to deal with an issue with Client Classification and it solved that problem. If you have the ability to go to 6.5 or 6.6 on a test AP, you could revert later if desired.
Photo of Mr B Williams

Mr B Williams

  • 3 Posts
  • 0 Reply Likes
Can I do this without upgrading HiveManager? If so, would you kindly share some instructions? 
Photo of Sjoerd de Jong

Sjoerd de Jong, Employee

  • 97 Posts
  • 20 Reply Likes
I would advise you to update the AP to 6.6r2a. 6.5r3a sometimes shows DHCP issues. My experience with 6.6r2a is that it's stable.
Photo of Tony Schaps

Tony Schaps

  • 28 Posts
  • 8 Reply Likes
I believe your HiveManager needs to be at at least 6.5 to have HiveOS 6.5 on your devices. That is not my area of strength, I will have to defer to others. 
Photo of Crowdie

Crowdie, Champ

  • 972 Posts
  • 272 Reply Likes
We experienced poor performance with AP121 and AP230 access points at one of our sites.  We resolved it by upgrading the APs to 6.5r4.   Upgrading the AP230s to 6.6r2 did not resolve the issue.
Photo of Dianne Dunlap

Dianne Dunlap

  • 75 Posts
  • 15 Reply Likes
I tested with a 230 on 6.6r1 HiveOS build2287 and got 391M with a single port connected.  The AP and PC were on the same segment with 1G/full negotiated on the same Cisco 3750 PoE switch.  The iperf client was linux.
You might want to snmp-monitor the AP and/or switchport to see what traffic levels.  If they are  nowhere near 1G and there is no correlation between performance and ethernet traffic, it may be more of an iperf issue than anything else.