High speed 5GHz network - 11n

From StarOS Community Wiki
Jump to navigation Jump to search
The printable version is no longer supported and may have rendering errors. Please update your browser bookmarks and please use the default browser print function instead.

This article needs more data and better formating

How to test

Excelent testing tool is program called "iperf".

On "server" side run: "iperf -s" or "iperf -D" to run it as an daemon (background)


On client side run: "iperf -c 10.0.55.5 -r -P 10". On 1GBps LAN it gave me from 845 to 849 MBits/sec. Test will run for 10 seconds with 10 simultanious threads.

"iperf -c 10.0.55.5 -r -P 10 -t 100" will run for 100 seconds, and

"iperf -c 10.0.55.5 -r -P 10 -t 100 -i 10" will run for 100 seconds, and report every 10 seconds.

Notice: "iperf -c 10.0.55.5 -u -b 100m" gives wild results like 41 Gbits/sec on 1GBps LAN so do not use it.

WAR-1a - WAR-1a, WLM200N5-26 2x2 HT40, 5Ghz,

Here are the results from a quick test using two WAR-1a systems. Note that we performed the test through the systems, between two large Intel i7 desktops. Connection Tracking on both WAR-1a systems has been disabled (important to achieve full throughput), and the network is routed.

Connection Tracking disabled, routed, v3 v1.5.16.2 PC (iperf client) ----x---- WAR-1a (Client) -- ( ~~25 ft~~ )-- WAR-1a(AP) ----x---- PC (iperf server)

2x2 HT40 = 88.4 Mbits/sec

WLM200N5-26 2x2 HT20, 5Ghz,

Connection Tracking disabled, routed, v3 v1.5.16.2 PC (iperf client) ----x---- WAR-1a (Client) --( ~~25 ft~~ )-- WAR-1a(AP) ----x---- PC (iperf server)

2x2 HT20 = 87.3 Mbits/sec

The following are between an Alix.3C (client), and a prototype AR7161 600MHz w/ GigE as the AP

both running v3 v1.5.16.2 Alix.3C (Client, iperf client) --( ~~25 ft~~ )-- Prototype(AP) ----GigE---- PC (iperf server) The throughput test was taken from the Alix console to circumvent the 10/100 Ethernet. All tests were done with Connection tracking being enabled on the AP, except for the last 2x2 HT40 test which has it disabled.

  • 1x1 11A HT20 = 50.2 Mbits/sec (from the Alix console)
  • 2x2 11A HT20 = 96.2 Mbits/sec (from the Alix console)
  • 1x1 11A HT40 = 112 Mbits/sec (from the Alix console)
  • 2x2 11A HT40 = 175 Mbits/sec (from the Alix console)

x4000 to x4000

[ 11] 0.0-10.0 sec 28.4 MBytes 23.8 Mbits/sec

laptop to pc on each side of exact same link

[ 3] 0.0-10.0 sec 21.9 MBytes 18.3 Mbits/sec

and

[ 4] 0.0-10.0 sec 31.7 MBytes 26.6 Mbits/sec

Just finished some new testing between an alix (single ethernet, 500MHz), and a war-1a (300MHz)

Test layout: PC (starutil client) --- alix (cpe) -- war-1a (ap) -- PC (starutil server)

WPA2 / PSK security in use

  • 11NA HT40+ 2x2: 11,100 KBps (88 Mbps) (limited by 10/100 ethernet)
  • 11NA HT40+ 1x1: 7,100 KBps (56.8 Mbps)
  • 11NA HT20 2x2: 6,900 KBps (55.2 Mbps)
  • 11NA HT20 1x1: 5,500 KBps (44 Mbps)

The throughput test does not increase the system load on either the alix or war-1a, both sitting around 1% (idle ssh session)

X86-PC(GigE ether) to X86-PC(GigE ether)

It is two IBM NetVista Towers, running the X86-PC Version, with Compex WLM200N5-26 Cards, and 30dBi DualPol Solid Dishes. It's a short link, under a KM.

Link shows 300Mbit, but sometimes drops to 270Mbit for a few seconds.

  • Performance is great - about 19,000 KBps (152 MBps) through this link with an HT40+ Channel, since the PC's have Gbit Ethernet, and that's true, uncompressed data, so that's very, very fast. That's almost 3x the throughput we got on an 802.11a Turbo link.

WRAP (266MHz)

WRAP (266MHz) caps out around 3,300 KBps (25.5 Mbps) and 90+ CPU load. Still not too bad considering their age.

XSCALE 266MHz (X2000) and XSCALE 533MHz test

Test layout: PC (starutil client) --- x2000 (cpe) -- war-1a (ap) -- PC (starutil server)

WPA2 / PSK security in use

11NA HT40+ 2x2: 4,200KB/sec (33.6 Mbps) 11NA HT40+ 1x1: 3,800KB/sec (30.4 Mbps) 11NA HT20 2x2: 3,200KB/sec (25.6 Mbps) 11NA HT20 1x1: 2,800KB/sec (22.4 Mbps)

System load scales proportionately with the throughput. HT40+ 2x2 is ~80 - 100%, HT20 1x1 is ~35%

XSCALE systems do not scale well with, nor designed for 11n cards, and do not perform as well as their successors.

UPDATE: XSCALE 533 (X4000) performs almost identically to it's 266MHz counterpart, with only marginal throughput gains.

UPDATE Feb, 07 2012: Gecko V4 should be able get 2-3 times more throughtput on XSCALE 533 and 266 systems. Testing is pending.


RouterBoard RB600 - RouterBoard RB600

Test layout: PC (client) --- 1 Gbps LAN --- RB600 (cpe) -- WiFI -- RB600 (ap) --- 1 Gbps LAN --- PC (iperf server)

11NA HT40+ 2x2: 113 Mbps (Both CPU's at 433MHz on 100%) 11NA HT20 2x2: 78 Mbps (CPU's on ~ 50-82%)


Maximum ranges

  • 11a: 36.29 miles (58.4 Km)
  • 11b: 65.64 miles (105.6 Km) (but limited to 32.82 (52.8 Km) via configuration, will be fixed soon)
  • 11g: 32.82 miles (52.8 Km)
  • 11a ht20: 36.29 miles (58.4 Km)
  • 11g ht20: 32.82 miles (52.8 Km)
  • 11a ht40: 18.14 miles (29.2 Km) (11a turbo is the same)
  • 11g ht40: 16.41 miles (26.4 Km) (11g turbo is the same)
Retrieved from "http://staros.tog.net/index.php?title=High_speed_5GHz_network_-_11n&oldid=2161"