Nv2-downlink-ratio - new huge bandwidth increase setting

RouterOS Wireless Networking
1

Sorry about this double post (the other post is in the Mikrotik Announcements section)
I think this post is important good news that all Mikrotik Admins should be aware of re v6.40

If as a Mikrotik admin you manage any Mikrotik nv2 links - you need to read below:

Re: Nv2-downlink-ratio

Whooo Ahhhh !!!
Wow !!!
Great !!!

I just tested Nv2-downlink-ratio (in v6.40) on one of my nv2 APs. I did NOT upgrade any clients. (Only the AP was v6.40 nv2)
I then performed some btests

#1 with the default Nv2-downlink-ratio = 50
btest UDP both — I get about 100 meg down to the client and 100 meg up from the client (symmetric up & down results during wireless saturation). This was the same result prior to upgrading the nv2 AP to v6.40

#2 with the default Nv2-downlink-ratio = 80 (this favors more bandwidth sent to clients at the cost of client slower uploads — during AP wireless saturation)
btest UDP both — I get about 180 meg down to the client and 27 meg up from the client (non symmetric up & down results during wireless saturation).

#3 with the default Nv2-downlink-ratio = 20 (this favors more bandwidth from to clients at the cost of client slower downloads — during AP wireless saturation)
btest UDP both — I get about 27 meg down to the client and 180 meg up from the client (non symmetric up & down results during wireless saturation).

#4 with the default Nv2-downlink-ratio = anything from 80-to-20 (((This test is performed when the AP is NOT wireless saturated)))
all speed tests results are normal bandwidth as previously experienced prior to v6.40 upgrade ((( When the AP in NOT wireless saturated )))

My conclusions:
A - Most normal NV2 APs to clients using the Internet average about-near-around 80 percent of all wireless traffic is from AP to clients.
In this scenario, setting the v6.40 AP Nv2-downlink-ratio setting to 80 can result in a whopping 80 percent more wireless bandwidth to customers during peak wireless saturation periods.
During peak-usage-periods when the nv2 AP is wireless saturated, you can sustain greater bandwidth from the AP to customers (+ 80 percent faster) at the expense of customer upload speeds (80 percent slower) during during peak-usage-periods when the nv2 AP is wireless saturated.

B - When the nv2 AP is not wireless saturated during lite-low wireless usage periods, the Nv2-downlink-ratio setting will not deliver any noticeable differences in wireless bandwidth to or from nv2 wireless customers.

C - I am guessing on this one — Saturated WDS/PTP links should also benefit using the Nv2-downlink-ratio on v6.40 nv2 links.
If the nv2 WDS/PTP AP is sending more traffic than receiving, then set the nv2 AP Nv2-downlink-ratio to 80 (to improve peak saturation bandwidth)
If the nv2 WDS/PTP AP is receiving more traffic than receiving, then set the nv2 AP Nv2-downlink-ratio to 20 (to improve peak saturation bandwidth)

D - Although ROS prior to v6.40 did not have a Nv2-downlink-ratio setting available - it behaves as if a Nv2-downlink-ratio is set to 50

E - If you want to maintain a symmetric wireless bandwidth for up & down during nv2 AP wireless saturation periods, then set the Nv2-downlink-ratio to 50


Summary:
This new Nv2-downlink-ratio setting in v6.40 is a super wonderful feature than should be able to much better handle nv2 AP wireless peak saturation bandwidths. This setting gives control to the Mikrotik admin to flavor/prefer/enhance/increase the wireless nv2 AP traffic during wireless saturation periods by allowing the Mikrotik admin to prefer up or down bandwidth when his nv2 APs are saturated during peak periods of the day.

Thank you Mikrotik

Note: I am also going to double-post this post into the wireless section of these forums. There may be many Mikrotik admins that do not follow this side of Mikrotik anouncements re this v6.40 topic

North Idaho Tom Jones

again - Thank you Mikrotik

Terrible NV2 Ac Network P2MP
2

Additional info:

In testing the new v6.40 Nv2-downlink-ratio on a test nv2 AP , I also discovered the following:

With Nv2-downlink-ratio already set to 80
I changed my TDMA Period Size from 2 to 3

I found some very favorable results on my very busy saturated nv2 AP (all tests performed on a saturated nv2 AP with 40+ distant connected clients

A test client customer connected was bandwidth testing prior to TDMA Period Size = 2 & Nv2-downlink-ratio = 80 at the following:
btest both udp 10s Tx 18-21 meg Rx 2.5 meg
btest send udp Tx 16 meg
btest receive udp Rx 8.2 meg

A test client customer connected was bandwidth testing with TDMA Period Size = 3 & Nv2-downlink-ratio = 80 now tests at the following:
btest both udp 10s Tx 76 meg Rx 11 meg
btest send udp Tx 58-87 meg
btest receive udp Rx 37-83 meg

By changing my nv2 v6.40 AP
from; TDMA Period Size = 2 & Nv2-downlink-ratio = 80
to; TDMA Period Size = 3 & Nv2-downlink-ratio = 80
I achieved btest speeds from the AP to the client at almost 400 percent faster throughput!!!

Question: Would somebody please check my results on their system and confirm or deny my results ? If this is confirmed, the average WISP might have the ability to increase average peak-usage/saturated wireless network speeds to customers anywhere from 100 percent to 400 percent.

North Idaho Tom Jones

3

another update:

Testing again - with over 70 distant clients with distance km ranges from 2k to 4k and testing to a 3k client
with the nv2 AP saturated with wireless traffic
I was able to btest at 10 to 40 meg to the test customer (during full saturation). (10 to 40 meg to the customer and about 5 to 15 from the customer) - using 3ms with 80 percent on the nv2 settings.

Now I know that anytime you run a btest on a wireless network, that you are effecting everybody on nv2 wireless network - and that everybody will suffer - however - 10 to 40 meg to the customer is pretty darn good with 70 distant customers on a nv2 wireless saturated AP.

North Idaho Tom Jones

4

Please make the test in tcp/ip

thank you Giuseppe

5

OK I just performed the same btest using TCP (not UDP this time).
Same customer I have been testing to - however this time we now at entering peak-usage times of the day

btest TCP send to customer 20-second average 21.4 Meg
btest TCP receive from customer 20-second average 15 Meg
btest UDP send to customer 20-second average 48 Meg
btest UDP receive from customer 20-second average 19 Meg

Prior to the v6.40 AP changes, this test customer was barely able to hit 10 UDP during peak usage periods (with 50 to 75 other clients on the AP at the same time)

I will learn more this weekend. I graph all traffic so that I can come back and compare/look at any 5-minutes/hour/day/week/month/year

There is a probability that I may need to change from 50/50 now 80/20 to something like 75/25 or 70/30 or 65/35 or 60/40 or 55/45 — Some testing will be needed

North Idaho Tom Jones

6

Hi Tom,

Thank you for doing this testing and publishing your results.

7

Hi TomjNorthIdaho,

first of all thanks for sharing your experience with us. and i see that you are testing point to multi point links, what about point to point links ? i test a little and see that nv2 higher ping then nstreme. do you have any experience about latency?

Thanks

8

Thanks for sharing your findings Tom! Cannot wait to do some testing.

9

Hi friends, could you tell me if the tests were done with dynamic or fixed downlink / uplink ratio. Thank you

10

fixed downlink

11

I’ve tested on Omnitik AC (set on A/N Ce channel) (15 SXT Clients connected, signals 45-60)
when fixed mode selected Ratio=60,70,80% throughput is worst than ever, I got better throughput with dynamic-downlink selected, but not as good as nv2 in v6.37.1
This one I could not understand why I have far better throughput with nstreeme, its like 100% better than nv2 (P2MP scenarios), what really sucks using 802.11 I got better throughput than nv2
I’ve also made test SXT SA sector, 10 clients connected I can tell you nstreeme can give up to 80-90mbits, nv2 max 30-40mbits TCP

In the last update only on P2P scenario fixed mode=80% NV2 Period Size=3 I got 15-20% increase of throughput from v6.37.1,

I really don’t know why nv2 protocol is such a disaster for me (speaking from my experience ), we use also ubq, airmax do what its suppose to do and don’t get me wrong I love mikrotik

its almost 4 years when nv2 come out but still almost on all P2MP AP-towers nstreeme protocol is selected…

12

@Lakis Same here NV2 & TCP speed test & streams = disaster , Nstreme & pure 802.11N Much much better.
New 6.40 ROS & NV2 fixed ratio 80/20 ratio worse than dynamic.
Cambium same hardware @ 170mbit/s DOWNLOAD TCP & UPLOAD 20mbit/s !.. from sector antenna p2mp 50 CPE @ 40mhz.

13

I tested on few nv2 sectors ROS v6.40
nv2 ratio 80 fixed vs 50 dynamic .
And all test is this some .

I not confirm downlink speed increase on this mode.

maybe in 40mhz channels ? (i use always only 20mhz )

14

Idaho Tom are your AP’s 802.11n or 802.11ac? I’ve personally noticed that NV2 performs HORRIBLE with AC protocol, have to use Nstreme but this new “fixed downlink” is only available in NV2 correct?

15

Most of my nv2 APs are N (N-only) 1x1
I only tested Nv2-downlink-ratio on N-only nv2 networks (with 24 to 65 client connections) all distant clients

I did test on an AC 2x2, but I do not remember the results.

I have not gone back to re-verify my results - yet.

FYI - if you can set your cell radius to 10 (or as low as possible), I find this often helps.
Also - I overclock my APs to the second-to-fastest cpu speed setting if they are not factory default at the fastest speed setting. I find this often help.
FYI - my nv2 APs are all strictly a WAN-to-vlan (wan-to-Ethernet) bridge. My aps are not doing NAT - not doing IP-dhcp - not doing firewalling - only AP and WAN of AP on a bridge.


North Idaho Tom Jones

16

dynamic downlink ratio 75% appears to working best so far but NV2 AP and NV2 PTP ( tried Nstreme but link kept disconnecting ) may not be ideal for PPPoE clients

17

What dynamic downlink ratio settings have you tried and what was the aprox difference in down/up speeds ?

I have not tested 75 yet.

North Idaho Tom Jones

18

Interesting information

Last week the day after v6.40 came out, I changed all of my nv2 dynamic downlink ratio settings on all of my APs

Today I was looking at a specific Cacti bandwidth graph servers which graphs a v6.40 client that makes a WDS nv2 link to my v6.40 AP.
This local wds client uploads to bandwidth to a remote AP tower I have which has 8 other nv2 APs for customers to connect to.
The remote tower wds AP has the “DownLink Ratio” set to 30 (I want to prefer traffic from the WDS client to the WDS AP.
I happened to notice the average CPU load on the WDS client dropped some after v6.40 and downlink radio settings and average bandwidth went up..
b4after.png
When I zoom in on my Cacti graph, it looks like pre-v6.40 CPU load was peak hitting 18 and now CPU peak load is hitting less than 12. This is measured at my busiest peak period of the day.

Interesting

EDIT: I also noticed the graphed heat sensor on the wds nv2 client dropped a little (and we have been experiencing hotter than normal days the last few days - I would of expected it to be hotter with the last few 100+ degree days)

North Idaho Tom Jones

19

TomjNorthIdaho less 5% CPU load is almost noting
in v6.40 Changelog
*) wireless - reduced load on CPU for high speed wireless links;

yes with nv2 downlink settings there is increase of throughput but only I can see benefit on P2P
on P2MP its awful.
I have made tests on 19km link (BaseBox5 Jurlios Antena A/N Ce channel signals 56/56)
When nstreeme protocol selected I got rock solid 115-120 mbit TCP (btest 5 connections)
8011.2 protocol 70-80mbits (btest 5 connections)
nv2 protocol downlink 75% period size=3 I got 60-126 mbit TCP, throughput always go up and down average is 75-85mbit (also made tests of all variants period size 2,3,4,5,6…auto, downlink 60-80%)
(btest 5-20 connections)

n21roadie@: when nstreeme selected please make btest only receive and only send see the difference. (Disconnect is possible one side collocation interference)

20

I try to minimize downtime for clients when attempting to tweak the network, the very basic tests will be check ping times and speedtest.net this is what customers use so must look at those results first with reference to a service package the test CPE is authenticated to, in other words if the package was 5/2 down/up - during peak usage hours what is the down/up speeds and ping times using PPPoE authentication, I also experiment with custom queue’s