Test the layer-2 and layer-3 Performance of a 24-port full Gigabit Switch

Performance Tests on layer-2 and layer-3 vswitches with 24 ports and all-gigabit vswitches have encountered many problems recently. Experts have said that QoS ensures port congestion, high-priority data is processed first. It makes sense to improve the utilization of 24 fully-gigabit switches with insufficient bandwidth.

Three-tier competition

Layer 3 24-port full Gigabit Switch testing includes OmniSwitch 6800-24, annett's AT-9924T/4SP, D-Link DGS-3324SR and SMC's TigerStack 8724ML324 port full Gigabit Switch. We tested the forwarding performance and QoS of Layer 3 and 24 full gigabit switches.

2-layer forwarding performance test

In this test, we use all 24 ports of the 24-port full Gigabit Switch for the full mesh test, we tested the packet loss rate, throughput, and latency of 24 full gigabit switches in 64, 512, and 1518 bytes. The testing time of packet loss rate and throughput is 120 seconds, and the delay test result is 60 seconds and 90% loads.

From the test results, the vast majority of 24-port full-gigabit switches have reached the goal of zero packet loss and 100% throughput per byte. Only the TigerStack 8724ML3 of SMC has 1518 packet loss at 0.12% bytes, the throughput reaches 97.7%. The test results are shown in table 6.

Layer-3 forwarding performance test

In the layer-3 forwarding performance test, the 24-port full-gigabit switch is configured with 24 VLANs, each of which includes one port, the IP address and subnet mask of each VLAN ranges from 1.0.0.1/255.255.255.0 to 24.0.0.1/255.255.255.0.

On the tester side, each port corresponds to a VLAN with a gateway address and an interface address, and each VLAN has four host addresses. We tested throughput, packet loss rate, and latency. During the test, all 24 ports of all 1-gigabit switches achieve 100% throughput and zero packet loss. The latency test results are shown in table 7.

QoS Test

QoS ensures that high-priority data is processed first when the port is congested. It is meaningful for networks with insufficient bandwidth to improve utilization. We tested layer-3 QoS by using the SmartBits 6000C tester to connect ports 1, 2, 3, and 4 of a 24-port full Gigabit Switch, ports 1, 2, and 3 send SNMP, POP3, and HTTP data streams that simulate UDP ports 161,110 and 80 to ports 4 at the same time.

The priority of the three data packets is the highest in the 24-port full Gigabit Switch, the highest in SNMP, the lowest in POP3, and the lowest in HTTP. The traffic on the three ports starts from 30% of the load and sends data packets to the congestion port. The total load on the three ports is 900 Mbps. Therefore, no packet loss occurs when the traffic on the three ports is 30%, then, starting from 30%, the load on each port increases by 10% each time until the load on each port is 100%.

The purpose of the test is to ensure that the packet loss rate of the highest priority is always zero during the test. The packet loss rate of the two low-priority packets should be different. In terms of data packet latency, the latency of high-priority data packets should be the lowest. According to the test results, each vendor has achieved the expected goal and the test results are satisfactory. Due to the limited space relationship, only one test result image is provided for readers to view. For the rest, see the website.

QoS test results: the green color indicates the high-priority packet loss rate of SNMP, the blue and yellow indicate POP3, and the low-priority packet loss rate of HTTP respectively. The high-priority packet loss rate should always be 0 during the test, total indicates the Total frame drop rate ).

In the QoS configuration method, the test engineer carefully observed the configuration file of each vendor, each vendor's QoS configuration is very distinctive, such as annett's AT-9924T/4SP When configuring QoS, first, data streams are differentiated based on certain conditions. Then, multiple data streams that require the same QoS policy are placed in the stream group, and multiple stream groups form a traffic level.

Execute some policies on the traffic level and place the policies to one or the global ports of the 24-port full Gigabit Switch. D-Link DGS-3324SR is more convenient when writing QoS commands, that is, the above actions are implemented in a command, but it is simple and has its disadvantages, write a policy for each level.

In this test, we made a comprehensive evaluation of 24 fully-gigabit switches. During the overall assessment, the weight is: performance accounts for 60%, functionality accounts for 30%, and price accounts for 10%. Currently, 24-port full-gigabit switches with fixed configurations are divided into two categories: Pure Layer 2 and Layer 3.

Alcatel OmniStack 6300-24 has outstanding performance in four fully-gigabit switches with 2-layer and 24-port tests. Alcatel OmniStack has excellent performance, including zero packet loss rate, 100% throughput, no HOL, and low latency; function: supports Protocol VLAN, layer-4 CoS, and so on. Based on the above factors, OmniStack 6300-24 won the online world evaluation lab editing selection award.