Test Method of IXIA1600 Tester for Gigabit Switches

The purpose of this evaluation is to make an objective comparison of each manufacturer's Gigabit Ethernet products. The main purpose of this comparison IS to provide the IS Administrator and other IT professionals with information that helps them make device procurement decisions. Therefore, our test is not limited to the performance test of the vswitch, but a comprehensive consideration, both using quantitative metrics such as throughput, packet loss, latency, and per-gigabit costs ), also use qualitative measurement criteria such as whether installation and management are simple and reliable ).

Our main test items are: configuration test-considering the flexibility, port density, and scalability of Switch configuration. Installation and usability testing-time and difficulty of installation, support documentation, and online help effectiveness. Feature testing: Includes Port Link aggregation, traffic control, MAC Address Table capacity, port image, VLAN, layer-3 switching, redundancy, MAC-based security, QoS, and Spanning Tree, multicast control. Management tests Test the capabilities of the console and command line interface, and support for Web, SNMP, and RMON. There are also important performance tests.

We use the industry-renowned network performance tester IXIA 1600 in performance testing. IXIA 1600 can be inserted with a maximum of 16 modules. Our test environment includes 5 10/100 M Adaptive modules, each of which has 4 10/100 Base-TX ports and 6 10/100/BASE-T adaptive LM1000T modules, each module has 2 10/100/M RJ-45 ports; 5 GBIC modules, each with 2 1000Base-SX/LX ports. Such a complete test environment allows us to test 12 1000Base-T ports, 10 1000Base-SX ports, and 32 10/100 Base-TX ports at the same time. Therefore, we can perform full load tests on the high-density gigabit port switches in the products to test their actual performance under the strictest conditions. During the test, we used Category 5 jumpers and optical fiber jumpers to connect the tested switch and tester.

The complete test environment allows us to test 12 1000Base-T ports, 10 1000Base-SX ports, and 32 10/100 Base-T ports at the same time. Able to perform full-load tests on the high-density gigabit port Switches of the products under test, and test their real performance under the strictest conditions.

In this switch test, we mainly use the ScripMate software of the IXIA1600 tester to configure and run various indicator tests. ScriptMate specially designed standard automation scripts for RFC 2544 and RFC 2285, we can easily define various parameters based on our own needs, while generating detailed log files and file describing results.

We have formulated nine testing indicators based on RFC2544, RFC 2285, And the Gigabit Ethernet test specification of China's communication industry, they are throughput, frame loss rate, back-to-back, latency, partial mesh, full mesh, back pressure, line-End Blocking, and error frame filtering, it basically covers the main performance indicators that need to be considered when you select a Gigabit Ethernet switch.

During the test, all ports of IXIA 1600 allow Adaptive Traffic Control and disable traffic control by default. In this test, three typical frames of 64-byte, 512-byte, and 1518-byte length are considered, unless otherwise specified, all tests are performed in full duplex mode. To ensure the reliability and accuracy of the test conditions, each test is repeated three times. The final result is to take the average value of the three tests.

In the throughput test, the port configuration is 1-to-1 ing, And the throughput is tested at full load. In the frame loss rate test, we set the initial speed to 100%, and test the frame loss rate through the ing between Port 1 and 1. In terms of latency testing, because a gigabit access switch includes a 10-Gigabit port and a 1-gigabit port, the latency between the 10-Gigabit port is significantly different from that between the 1-gigabit port, therefore, we conducted three tests on the same module, cross-module, and Gigabit ports of Mbps. Each test selected a pair of two-way data transmission ports, for a vswitch with an abnormal delay of 100% wire speed, we will test the speed adjustment slightly lower. In the mesh test, a full mesh test is performed on a gigabit backbone switch. For a gigabit access switch, a partial mesh test method is used to match each gigabit port with 10 10-Gigabit ports, the remaining MB ports are used for full mesh testing. In the Back-to-Back test, the port configuration is 1-to-1 ing under full load, and the initial speed is set to 100%. Back-pressure testing uses two methods. In half-duplex and full-duplex states, data is sent to one port through three ports to check whether back-pressure and IEEE 802.3x traffic control are supported. For line congestion, port A and B are used to send data to port C to form A congestion port, while port A also sends data to Port D to form A non-congested port. For error frame Filtering, we use one-to-multiple ing to test the six typical error frames, namely, small frame, ultra-large frame, CRC error frame, fragment, alignment error, and dribble error.

During the configuration, installation, and usability, features, and management tests, our engineers carefully read the instructions provided by the vendor, and actually configured and installed vswitches, the Console port, Telnet, and Web are used to configure and try vswitch management methods, link aggregation, VLAN, MAC address filtering, static Unicast address, port image, and other functions, we managed the vswitch using the network management software provided by various vendors. In addition, we also used a MIB Browser to view the MIB information of the vswitch.

In our overall evaluation of a Gigabit Switch, performance accounts for 40% and cost-effectiveness accounts for 10%, while other functions, configurations, features, management, and ease of installation account for 50%.

1. Introduction to nine metrics for performance testing of Gigabit Switches