Last month there was a post to a thread from Mark McKibbin (see
http://lists.wispnotes.com/pipermail/karlnet/2003-October/003153.html
<http://lists.wispnotes.com/pipermail/karlnet/2003-October/003146.html>
) in regards to the KN50 "Blackbird" units. His questions were in
regards to the poor link quality of the Orinoco Gold cards and other
cards that have been included with the units. We have been experiencing
similar problems (our units were purchased with Dell TrueMobile Radios
(Agere chipset Firmware rev. 8.10.1)). I have been running tests on a
number of different cards with the KN50. The following are the results:
NOTE: These tests were conducted on live point-to-multipoint
connections. We used all the same hardware for our testing; the only
change was the radio card for each test. The KN50 board was classified
as a SOHO CPE with product v1.00-00-073017. Our average connection
speed with a 1024k key is around 700-900k. The Base units we used for
testing are using software versions 4.43 and 4.44. They have a varied
number of subscribers, from 4 (a new site) up to 50+ (our oldest site).
On all these tests, the SNR was between 12dB and 18dB upstream to Base
and 21dB to 30dB downstream from Base.
Dell Radio:
Chipset: Agere
Firmware ver.: 8.10.1
PING test (a standard IP tools PING test from a PC through the Base to
the CPE (100 packets)): 73% of the packets lost
Throughput test (5 bandwidth tests from 2wire.com and dslreports.com):
278, 112, 245, 237, 298 (speed measured in kilobits)
NOTE: The packets lost in the PING test are RARELY noted on the client
side (i.e. surfing the Internet, Web Page comes up unavailable). They
just notice a significant slowdown.
Orinoco Gold:
Chipset: Agere
FW ver: 8.72.1
PING test: 64% lost
Throughput: 451, 368, 519, 259, 311
Orinoco Silver:
Chipset: Agere
FW ver: 6.6.1
PING test: 0% lost
Throughput: 676, 574, 481, 625, 616
Avaya Radio:
Chipset: Agere
FW ver: 6.8.1
PING test: 0% lost
Throughput: 550, 447, 380, 571, 561
Wave Wireless Radio:
Chipset: Agere
FW ver: 6.8.1
PING test: 1% lost
Throughput: 515, 660, 674, 553, 622
Demarctech Prism Radio:
Chipset: Intersil (800C)
FW ver: 1.3.6
PING test: 3%
Throughput 591, 745, 652, 733, 725
At the location where we were doing the final throughput tests, an NDIS
install in a local server has the following results:
PING test: 1% lost
Throughput: 725, 989, 867, 967, 802
I realize that the PING test is hardly conclusive, but it definitely
coincides with the lower bandwidth connections. Does anyone else have
similar findings? One other point of note is that the PING test on a
unit connected to a highly utilized Base has significantly more dropped
packets on the Dell and Orinoco Gold cards (i.e. The cards are noted
with a 73% and 64% loss on a Base with 40+ subscribers, these numbers
drop to 35% and 23% on a Base with 20 or less clients). This would
indicate Base unit over saturation, but we can connect a standard NDIS
(radio) client or a KN100 board and the PING tests and throughput tests
are excellent (0% loss, avg. 880k throughput).
It appears that cards running older versions of firmware (6.x.x) do not
have the problems that newer firmware (8.x.x) have; at least in regards
to the KN50. I have upgraded the Orinoco Silver card to v8.7.5 firmware
and it exhibits the same problems. I downgraded the firmware and the
problem goes away. Our only option, at this point, is to downgrade the
firmware on the newer cards or try to find older silver cards with the
external antenna attachment (seeing as how the new silvers do not have
them anymore).
I tested the Orinoco Gold and the Dell Radio in a KN100 board with the
following results:
OG:
PING test: 0% lost
Throughput: 714, 784, 987, 966, 908
Dell:
PING Test: 1% lost
Throughput: 837, 664, 843, 767, 813
This looks like a problem with the KN50's and newer Agere firmware.
Anytime you install 8.x firmware on any radio in the KN50, performance
is significantly reduced and there is a high amount of IP packet
failure. Our research team would like to know if anyone else concurs
with these findings and is there a solution in the works. If not, is
there anymore testing you would like from our end to get this issue
resolved?
Tim Steele & Marcus Rudd
Wireless Networking Engineers
e-ISCO Internet
www.e-isco.com <http://www.e-isco.com/>
_______________________________________________
Karlnet mailing list
Karlnet@WISPNotes.com
http://lists.wispnotes.com/mailman/listinfo/karlnet
|