27 clients. Mixture of Mainly SXT-L, one SXT-Lac and one SXT-sq5ac and rb911 and some 711's and even still 3 411's.
AP in bridge mode, CPE's routing with PPPoE on wlan1
Management network by dhcp server on RouterBOARD 750G r3 behind AP with vlan2
vlan2 interface on CPE's wlan1 with dhcp client.
Tests are done by Mikrotik's bandwidth test from CPE with 1 connection only, tcp and download.
40Mhz channel width
AP running ros v6.42rc52 and fw updated to 6.42rc52:
I left not relevant items out.
Code: Select all
adaptive-noise-immunity=ap-and-client-mode band=5ghz-onlyn basic-rates-a/g="" channel-width=20/40mhz-Ce disconnect-timeout=3s distance=dynamic frame-lifetime=0 guard-interval=long ht-basic-mcs=mcs-7,mcs-8,mcs-9 ht-supported-mcs=mcs-7,mcs-8,mcs-9,mcs-12,mcs-13,mcs-14,mcs-15 hw-fragmentation-threshold=disabled hw-protection-mode=none hw-protection-threshold=0 hw-retries=3 interworking-profile=disabled keepalive-frames=enabled l2mtu=1600 mode=ap-bridge nv2-cell-radius=10 nv2-downlink-ratio=75 nv2-mode=dynamic-downlink nv2-noise-floor-offset=default nv2-qos=default nv2-queue-count=2 nv2-security=enabled on-fail-retry-time=100ms preamble-mode=long rate-selection=advanced rate-set=configured rx-chains=0,1 security-profile=AP-E supported-rates-a/g=12Mbps tdma-period-size=3 tx-chains=0,1 wireless-protocol=nv2 wmm-support=enabled
AP can be switched between wireless-protocol=nv2 or 802.11
CPE's are all set accordingly and all run on v6.41.3 and fw 6.41.3. Only SXT-sq5ac is running on 6.42rc52 too with latest fw. (6.41.3 doesn't support PPPoE yet for 'arm' device!) :
Most other setting are by AP. RTS/CTS is set to work always in 802.11 mode.
Code: Select all
adaptive-noise-immunity=client-mode band=5ghz-onlyn basic-rates-a/g=6Mbps bridge-mode=enabled channel-width=20/40mhz-Ce compression=no default-authentication=no disconnect-timeout=3s distance=dynamic frame-lifetime=0 frequency-offset=0 guard-interval=long hw-fragmentation-threshold=disabled hw-protection-mode=none hw-protection-threshold=0 hw-retries=3 interworking-profile=disabled keepalive-frames=enabled l2mtu=1600 mode=station mtu=1500 multicast-buffering=enabled multicast-helper=default nv2-security=enabled on-fail-retry-time=100ms preamble-mode=long security-profile=AP-E tx-chains=0,1 wireless-protocol=any wmm-support=enabled
This network runs in an area where basically every frequency has some other Wifi AP working. Hence interference but the chosen channel is as free possible.
On a live network but little usage, I open 4 clients.
From each CPE I open a single stream tcp download from the rb750gr. In this network no queues or priority settings in place.
so the stream runs from the rb750gr over a gigabit cable though a Netonix switch by gigabit to the AP and then by wireless towards the clients.
All clients connect basically in a -38 to -55 range and the worse is -57
Clients for the test all run with low 50's or better.
Test 1. 802.11 'n';
When one client start, 80 to almost 100mbps download. When second client start 1st looses some but both on sort of 30-40Mbps. Add no. 3 and no. 4 each individual speed goes down somewhat further but still they all fluctuate between 20 and 30Mbps.
Meanwhile the AP shows a throughput running up to a 115mbps average with some peak of 130Mbps.
All 2 minutes test.
Test 2. NV2, variable downlink 75%:
One client starts with some 40-50mbps, peaks up to 70 max. When more clients are add, each get less. In the end they all four run with only some 10-20mbps.
Meanwhile the AP shows a throughput running up to a 70mbps average with some peaks nearing 115mbps. But overall much more unstable as in Test 1
Test 3. NV2, fixed downlink 75%:
Sort of same as Test 2 but everywhere and overall we lost another 5-10% on capacity.
My conclusion: 802.11 'n' beats NV2 even on a multiple CPE high interference network.
To be honest, in preparation of a swap of 3 other networks that now run Mimosa we tried first with SXT's ac's in Mimosa 'interop' mode which is in fact plain 802.11 with fine tuned RTS/CTS and we were already blown out of our shoes by the immense boost in speed we got on the clients. I partially blamed that on the more powerful Mimosa AP combined by the more powerful SXT-ac's but after reading some posts by some guys somewhere on the net that do a lot of testing I already got the impression that the 802.11 with RTS/CTS isn's to bad as what every vendor makes us believe. (They all try to talk us into their proprietary tdma protocol and if succeeded less change a client will walk to the competition again....
My next project will be now to exchange all 'n' units in this 'test' network by 'ac' mode CPE's which is going to take a month or so.
In the mean time Mikrotik should do its best to work even more on their latest NV2 improvements!
When all my network is 'ac' I hope to come close to what we have with Mimosa now, 100, 150Mbps or even 200mbps to the client! When this can be achieved I will be back supporting 100% Mikrotik.
Because after all, a 30 client P2MP 'ac' network of Mikrotik is still only costing a fraction of what eCambium, Mimosa are costing. We are still cheaper then ubnt.....