Appendix A - Performance Testing

Scope
The scope of these tests are to evaluate stability and performance of Linux, driver, and wireless network interfaces using various network configurations to achieve optimal performance. These tests do not account for interoperability and compatibility of various wireless NIC manufactures.

Description
The following is a description of the test tools, test cases, configurations, and assumptions to perform performance testing of the target system. This document is accompanied by diagrams detailing the test case configurations.

 

TARGET TEST1 TEST2

 

TARGET
The target platform is Pcengines Wrap 1D GEODE 266 Mhz with 128 MB memory with 2 VIA VT6655 802.11g network interface cards and 5dBi omni antennaes extended by a 14cm IPEX to RP-SMA antenna assembly.

TEST1
Laptopi Windows XP using Broadcom a wireless NIC with network performance test software NetIQ.

TEST2 Alternative
Embedded VIA Eden 800 MHz Linux using a VIA VT6655 802.11g wireless NIC with network performance test software netperf.

Notes and Recommendations
Based on observed results it is recommended that all interfaces are configured with static (manual) addresses. While the dhcpi leases are initial set to 24 hours there is a concern that if there is achange in lease addresses if this would disrupt the testing software. For example, if the test server initially receives an address of 192.168.55.40 and the test client receives an initial address of 192.168.77.40 then traffic is pushed from one platform to the other using these addresses. If anytime during the testing a lease expires and a new address is assigned then the test software would have an invalid endpoint.

The recommended channel settings are to such that there is maximum frequency seperation. It is also recommended that a site survey is conducted such that there are no rogue or interfering devices in the test area.

With regard to Legend L3 for TC5, the ultimate goal is to utilize a 5Ghz channel and utilize IPv6 autoconfiguration on all joining wireless mesh interfaces that provide backhaul services to the nearest WAN uplink. IPv6 is not within the scope and 5Ghz radios and 802.11ai network interface cards are not available for this test.

The recommended number of target system is three (3) to reduce setup time. The default configuration is TC5. Preconfigure one (1) system for TC1 to TC3 and use the other two (2) systems for TC4 and TC5.

For system #1 preconfigure the WWAN interface using the above reference for TC1 to TC3 Systems #2 and #3 should boot into TC5. To issue TC4 simple perform the following commands on each of the systems console as follows:

System #2
cp /etc/pipe/examples/side-a-pipe.conf /conf/pipe.conf
/etc/rc.pipe start

System #3
cp /etc/pipe/examples/side-b-pipe.conf /conf/pipe.conf
/etc/rc.pipe start

 

X = 22Mbps???

Test Case 1
Purpose: 2 client single interface performance, driver evaluation, radio resource contention
Setup Time: 5 minutes
Duration: 24 hours
Throughput Expected: ½X

Test Case 2
Purpose: Dual interface performance
Setup Time: 10 minutes
Duration: 24 hours
Throughput Expected: 1X

Test Case 3
Purpose: Interface performance
Setup Time: 10 minutes
Duration: 24 hours
Throughput Expected: 1X

Test Case 4
Purpose Bonding two network interfaces to achieve a dedicated load balanced link redundant wireless pipe with about 2X performance
Setup Time: 10 minutes
Duration: 24 hours
Throughput Expected: 1¾X

Test Case 5
Purpose: Ad-hoci mode testing
Setup Time: 10 minutes
Duration: 24 hours
Throughput Expected: 1X