throughput issue w/ ZF 7962

  • 1
  • Question
  • Updated 3 years ago
I just stumbled on to a weird problem that's completely baffling me... I have a ZoneDirector 1000 controller and three ZoneFlex 7962 devices. One of the ZF 7962 is exhibiting very slow throughput - given that our office's connection is 100d/20u and that the other two ZF WAPs perform as expected when running speedtest.net, this third unit shows 15d/10u and I can't figure out why.

I've performed speed tests, via the ZoneDirector, to each of the units and they max out, so it doesn't appear to be a wiring issue. Even when I join my laptop to the WAP in question and I'm literally standing ten feet away from it, my performance speeds stink.

Any troubleshooting tips/recommendations?
Photo of AgencyPJA

AgencyPJA

  • 8 Posts
  • 0 Reply Likes
  • frustrated

Posted 3 years ago

  • 1
Photo of Primož Marinšek

Primož Marinšek, AlphaDog

  • 413 Posts
  • 48 Reply Likes
If you used the integrated ZAP in the ZD and it shows max tp on cable WITHOUT any packet loss than it may mean that you have interference in one location.

You can maybe check the packet retry rate at the AP and post it here so we'll get a better clue.

Go to the Monitor :: APs in the ZD and look for "Histogram of PHY errors per second". If there is alot of ph errors from 2k up then that's most likely the case.
Photo of Michael Brado

Michael Brado, Official Rep

  • 2182 Posts
  • 300 Reply Likes
Primoz is right, if that AP is more affected by interference in the client band than on the other two. The System Info file (click first icon to the right of AP under Monitor/
Access Points page), is a text file you can read with Notepad++. Athstats are the
radio details (radio0 = 2.4, radio1 = 5G), and the Histogram shows a 2 minute
rolling RF sample. This data appears at the bottom of the files, just before AP
syslogs.

Secondly, since most modern clients can use all 11 of the 2.4 channels, you may
wish to employ the ChannelFly algorithm to manage channel assignment, (from
the Configure/Services, Self-Healing section). Having 3+ APs is helpful, and
ChannelFly also looks at non-802.11 interference, to help select the best of all
11 available 2.4G channels. You should see optimal throughput after Channelfly
has had a little time for your APs to build their internal channel/interference tables.

Lastly, compare Speedflex from the ZD to APs/clients, as well as your othe client
speed tests. You may find good local network times, but that your Internet
connection is the slow point.
Photo of AgencyPJA

AgencyPJA

  • 8 Posts
  • 0 Reply Likes
First off, thank you both for your input.

The Speedflex test for allthree WAPs show both up- and down-link to be 400+Mbps w/ zero packet loss.

As for the Histogram, here's what I've got:

(rafters) Histogram of PHY errors per second (pcttime in each range)
0 1-500 ..1K ..2K ..5K ..10K ..20K ..50K .100K more
0 1 2 26 57 12 2 0 0 0

Compare that to my two other WAPs:

(annex) Histogram of PHY errors per second (pcttime in each range)
0 1-500 ..1K ..2K ..5K ..10K ..20K ..50K .100K more
0 3 18 21 31 22 5 1 0 0

and

(mezz lng) Histogram of PHY errors per second (pcttime in each range)
0 1-500 ..1K ..2K ..5K ..10K ..20K ..50K .100K more
0 3 29 45 17 2 2 1 0 0

And I don't see much of a difference...

I've turned on the ChannelFly function for both 2.4 and 5; will report back in a few days.

Thanks again...
Photo of Michael Brado

Michael Brado, Official Rep

  • 2182 Posts
  • 300 Reply Likes
All three APs show higher than desired number of PHY errors. You don't want to
see over 2K/second. The first two APs report 60% of packets have 5K+ PHY errors
per second.

(rafters) Histogram of PHY errors per second (pcttime in each range)
0 1-500 ..1K ..2K ..5K ..10K ..20K ..50K .100K more
0 1 2 26 57 12 2 0 0 0

Compare that to my two other WAPs:

(annex) Histogram of PHY errors per second (pcttime in each range)
0 1-500 ..1K ..2K ..5K ..10K ..20K ..50K .100K more
0 3 18 21 31 22 5 1 0 0

and

(mezz lng) Histogram of PHY errors per second (pcttime in each range)
0 1-500 ..1K ..2K ..5K ..10K ..20K ..50K .100K more
0 3 29 45 17 2 2 1 0 0

These histograms indicate some possible constant EMF interference near
the APs. You don't want them too close to florescent lighting ballasts, electrical
conduits, heating/AC ducts, elevators, machinery, etc.

Have you run Channelfly on 2.4G for a little while to see if these numbers might
come down after letting the algorithm evaluate interference?
Photo of AgencyPJA

AgencyPJA

  • 8 Posts
  • 0 Reply Likes
Turned on ChannelFly just this morning; I'll give it a few days and report back.
Photo of Primož Marinšek

Primož Marinšek, AlphaDog

  • 413 Posts
  • 48 Reply Likes
If you connect to the 5G band are things just as bad?
Photo of AgencyPJA

AgencyPJA

  • 8 Posts
  • 0 Reply Likes
So, I've been running the APs w/ ChannelFly for a week now and I'm sorry to say I'm seeing no change in the PHY errors.

It's worth saying that the units seem to perform fine (apart from the fluke that caused me to report this in the first place) - it's likely that these units have had these interfrerence issues all along but it doesn't seem to affect throughput, as far as I can tell.

Am I correct in thinking that connecting to 2.4 or 5GHZ band isn't the issue, that the reports are reflecting general interference issues?
Photo of Michael Brado

Michael Brado, Official Rep

  • 2182 Posts
  • 300 Reply Likes
The radio Athstats Histogram is a rolling 2 minute sample of RF.

If you have an AP support info file taken when your test clients are connected,
you can also analyze the client RSSI (receive signal strength indicator) and PER
(packet error rate) values. A low RSSI indicates client distance from AP and is
directly proportional to the data rate. PER indicates the client is near to a source
of interference, causing local packet drops.

A "good" client connection will have PER < 10% and RSSI - PER connection, and might result in the lowered speedflex/speednet test
results.
Photo of Michael Brado

Michael Brado, Official Rep

  • 2182 Posts
  • 300 Reply Likes
The above reply was cut off by the browser/editor...

A "good" client connection will have PER < 10%, and RSSI - PER <=25.

Here is an example client connected on channel 1 on 2.4G

-- wlan0 --
ADDR AID VLAN CHAN RSSI IDLE CAPS STATE HTCAPS ADMITTEDBW
1c:ab:a7:d8:2a:ac 1Y 1 1 25 0 ESS PRIV SHTPRE QOS ERP HT PWRMGT OWL ST-PS RXSTBC 0 RSN AES AES PSK WME

STA: 1c:ab:a7:d8:2a:ac
rx_data_frm 240634 rx_mgt_frm 8592 rx_bytes 33887185 rx_dup 635 rx_wepfail 1
tx_data_frm 866347 tx_mgmt_frm 8550 tx_bytes 489598655
tx_assoc 1 tx_auth 1
good_tx_frms 874897 good_rx_frms 249226 tx_retries 293563
tx_rate 24000 tx_kbps 18023 rx_crc_errs 36299
tx_per 16 ack_rssi 21 rx_rssi 18

This client has an rx_rssi of 18, indicating a far distance from the AP, resulting in
maximum data rate of 24mb. But a PER of 16% means they are dropping packets
due to interference near the client. This PER can result in lower speedflex/speednet
test results.
Photo of Primož Marinšek

Primož Marinšek, AlphaDog

  • 413 Posts
  • 48 Reply Likes
Hey Michael

Thanks for the answer

When you say that the client has rx_rssi of 18 you are talking about SNR right not some arbitrary number here?

And the tx_per is for packet error rate or something else? If it's so, what's the unit there and what number is an indication I need to check something?