Thought I'd share my test results on NStreme.
I am happy with John Tulley's (Mikrotik's) attitude, in the sense that they
are constantly trying to improve their product, and listening to comments
from users of their product, and attempting to make sure users are not
misinformed about the features of their product. They are smart guys, that
have a lot of skin in the game, and experience beyond many in this industry.
I think Mikrotik is on to something, in their attempt to make a better
protocol to enhance Wifi, and commend them on their guts to do so. However,
with that said, I was rather disappointed with the results of NStreme in my
testing. I think its still needs a lot of work.
I used the same test bed I've been talking about the past few days
(originally with Trango), 10 miles, -47db rssi on Mikrotik w/Range5, one
radio per 532 MB, 2 ft quickfire dish.
My goal was to test how Mikrotik handled packet loss, when it was thrown at
it at different speeds. I used three primary tools for testing performance.
Mikrotik's included Btest/bandwdith tester and Iperf TCP and IPerf UDP.
I was surprised to see that Mikrotik's Built-in BTest program actually
performed (on Average) pretty darn close to the results that I got with
Iperf. It was hard to tell that at first because BTest is a bit jumpy with
sparatic speeds, but the average reading was pretty close if an eye was kept
on it. Iperf was more accurate in getting precise results. The most useful
test is Iperf UDP. The reason is that Iperf will show you at what point
(speed) you start to get packet loss and how much. All of Mikrotik's
performance tests, leave out "packet loss" in their results, so you can't
see the effect or choke points. In a real world enviroment, with lots of
subscribers and over subscription, its likely that a link will get hammered
from time to time, and nice to know what will happen from a packet loss
point of view when limits are reached. But the biggest help of Iperf UDP,
is to detect the MAX speed possible from the radio, and at what trade off of
packet loss. The Iperf UDP speed results is what should be utilized for
configuring bandwdith management tools. Tke note: That Iperf purpose is to
see what point packetloss will occur, and it ALWAYS occurs with any
connection which is pushed beyond its limits. no inteligence at other layers
is applied to slow tranmittingto reduce packet loss.
My goal was to compare Nstreme to not using NStreme. (used new version 9.12)
The majority of the time the Station side, negotiated at 54 mbps, however on
the AP-bride side negotiated speed could jump around from 48mbps to 6 mbps.
but usually around 36 mbps (QAM16) majority of the time. I was watching the
negotiated rate at the same time as testing to look for modulation change.
It didn't change often. To be clear tests were done with WDS mode, and all
default auto settings. Polling was turned off since a PtP link. And best
fit, was tried on and off, however, that would not have much effect, as
Iperf was sending consistent size packets of 1470 bytes. Often people will
see better speeds and less packet loss when testing with smaller packets,
but to get menaing ful results its important to test a full packet size.
Mikrotik reported a Tx CCQ of 93-98% and noise floor of -101, in status of
WLAN..
In Iperf you set the speed at which you want to send data to the link, and
then it reports the speed transfer at and at what packet loss. I ran the
tests several times for each speed, so I had a good average to consider.
I was stunned by the results.
Using NStreme....
Throughput (mbps) / Packet loss.
6M / .12%
8M / 1% - 2.4%
10M / 1.2% - 8%
12M / 4%
13M / 8.3%
15M / 21%
18M / 35%
In summary, NStreme could perform well without packet loss at about up to
8mbps.
Using Standard Wifi....
10M / .73%
12M / .36%
13M / 1.8%
18M / 2.2%
19M and up- started to see high packet loss
Without NStreme, I could push almost 18mbps at the same packet loss
Nstreme's 8mbps. And at 12mbps, I got very low packet loss. So in summary,
Standard Wifi doubled the throughput of NStreme.
Unless there is some hidden tuning commands for NStreme, its not cutting it
yet, over default Wifi.
Using Mikrotiks BTest, I got about 8M (4mbps in each direction) with
NStreme, and about 12M (6-7mbps one way, and 5-6mbps in the other.).
Likewise I tried Iperf TCP, which produced results very similar to
Mikrotik's average.
Note: understand that this enviroment may have some noise considerations,
tested to be around -80db with Trango. I tested the Mikrotik using 5.3Ghz,
but performance was not better, but 10.5 miles is a long shot for 5.3Ghz at
the lower radio output power.
If anyone has any hints on improving NStreme performance let me know, I'll
have the test link up for about another 24 hours.
Tom DeReggi
RapidDSL & Wireless, Inc
IntAirNet- Fixed Wireless Broadband
--
WISPA Wireless List: [email protected]
Subscribe/Unsubscribe:
http://lists.wispa.org/mailman/listinfo/wireless
Archives: http://lists.wispa.org/pipermail/wireless/
