Through a site router, yes, that's exactly what I was seeing.
On 3/6/2016 4:19 PM, Ken Hohhof wrote:
Are you saying this is an upstream problem? I don't see how what you
describe can happen with downstream traffic if you have a tower router
with interface buffers.
-----Original Message----- From: George Skorup
Sent: Sunday, March 06, 2016 3:42 PM
To: af@afmug.com
Subject: Re: [AFMUG] Poor throughput on ePMP AP
What about sync/GPS? And are you on 2.5 or 5ms framing?
Nate @ Blast and I had a similar issue. We had 450 + Ubiquiti + Trango
ApexPlus. Throughput was fine on the PMP450. Directly attaching to the
switch/router at the Trango link showed no apparent issues. Across a
Rocket link, the clients behind it (some UBNT, FSK and 450) would see a
burst of 10-12Mbps and then it would dwindle to 3-5Mbps. We swapped the
Rockets for a set of ePMP. Throughput got much better with the link in
ePTP (2.5ms) mode. So my first thought was that the Rocket link just
sucked, and it did, but we were still getting some speed complaints and
not even at peak utilization. Something else was going on.
And then Nate posted about very similar issues. He had a client on a 450
and an ePMP. Throughput on the 450 was fine, ePMP not so much. And
really only when he had the traffic going across his Trango link. That's
when Gino mentioned the ApexPlus ethernet packet buffer issue. The
shorter frame duration would allow the Trango buffer to empty a little
faster, hiding the problem just a little bit. Ultimately, Trango needed
to get their buffer issue sorted out, and they did.
I wrote all that to ask this... what's farther upstream from this site?
On 3/6/2016 10:26 AM, Justin Wilson wrote:
Sp I have an issue. Tower is a 4 sector ePMP setup and two
backhauls. We are seeing poor throughput when connected to the APs
(all of them). Signals are great. Quality and capacity are great.
When going through the AP we are consistently seeing 4-5 megs of
download out to a known speedtest server. If we plug into the same
wired port on which the AP, which gave us the poor throughput, we can
max out the 100 meg uplink. Issue only happens when going through the
wireless. Here is what I know:
1.APs are set to 75/25. SM can do a link test and get 50meg x 12meg
on the link test. So RF is good. Isolated AP to where only one
client was on. Same great results.
2.Firmware does not seem to be a factor. Can reproduce this on
2.6.1, 2.5.1, and 2.4.3.
3.Have replaced POE injectors 3 times with different manufactures.
4.Does not matter if it is DHCP or PPPoE. Hard wired is fine.
Wireless stinks.
The AP is plugged into a POE which then plugs into a Mikrotik 2011.
If I plug into the same exact port the aps plug into speeds are
great. The only thing that is left is patch cable to POE, Cable to
AP, or AP. I refuse to think 6 cables on the tower do the exact same
thing. The odds for that are Powerball winning crazy. Speediest from
a laptop hooked to an SM to the 2011 are poor. 4-5 megs consistent
with spikes up to 10-13, but all over the place. Acts like
negotiation. Have set Mikrotik to Auto, to 100 meg full, only
accepted 100 meg on auto.
Customers who have 5 meg packages or below don’t see this. Those with
10 meg packages are the ones seeing 4-5 meg speeds. 10 meg packages
did work.
Clients are mainly at 2.6.1, but we can reproduce this with a 2.5 and
2.4.x SMs. Nothign has changed in the network, which tests out just
fine up to the point it gets handed off to the ePMP.
Anyone ran into this? Any thoughts?
Justin
