How we did it
 
|
|||
 | |||
|
We tested the Cisco Catalyst 3550-24 running Cisco's IOS Version 12.1(6)EA1a. We used Spirent Communications' Smartbits 6000 with 24 Fast Ethernet ports and two Gigabit Ethernet ports to generate traffic for Layer 3 throughput, route convergence and quality-of-service tests.
For the Layer 3 throughput tests we configured the switch for forwarding packets at Layer 2 and Layer 3. We then ran a full mesh test using the throughput and latency tests in SmartFlow using 64-, 128-, 256-, 512-, 1024-, 1280-, and 1518-byte packet sizes for all tests. We then observed the frame loss rates and latency values.
For our route convergence tests we rebooted the switch to clear all memory at the beginning of each of these tests. For the route capacity test we then used Spirent's TeraRouting Test on only the gigabit ports to advertise routes. IP routes were advertised iteratively in a binary search fashion to determine the route table capacity. We repeated the test for Routing Information Protocol (RIP) and Open Shortest Path First (OSPF).
For the route convergence test we allowed traffic to flow in a bidirectional manner through the two gigabit ports at 90% wirespeed load. We advertised 80% of the maximum number of routes supported while sending traffic flows over those routes at 90% of wirespeed. We then withdrew 50% of the routes after a configurable amount of time and then re-advertised those routes. We then observed and graphed the frame loss rates for the traffic flow. The test was repeated for RIP and OSPF.
For our QoS testing we ran three sets of tests.
In our rate-limiting test, we configured a QoS access control list (ACL) on the switch to match up any User Datagram Protocol (UDP) SNMP traffic. We assigned a rate-limiting policy of 50M bit/sec to each port. We then used SmartFlow to send traffic at rates of 10% to 100% of wirespeed in 10% increments and observed the rate-loss patterns.
For out ACL performance test we created an ACL that permitted all SNMP traffic to flow through the switch, applied to five, 10, 15 and 20 ports.
For the final QoS test we configured SmartFlow to send DSCP traffic from 16 Fast Ethernet ports to the remaining eight Fast Ethernet ports. We configured each ingress port to send eight flows of traffic with DSCP values of zero, eight, 16, 24, 32, 40, 48 and 56. We ran the last test with four iterations using traffic rates ranging from 25% to 100% wirespeed. The expected result is for the flows with DSCP values of zero and eight to have the highest latency (lowest priority); 16 and 24 to have the next highest latency; 32 and 40 to have the next highest latency; and 48 and 56 to have the lowest latency values (highest priority). We repeated the above test with the same 3550 configuration while sending a larger amount of traffic that would cause the egress port to be congested and measured packet-loss rates.
RELATED LINKS
Bass, a senior technical staff member at CNL and co-author of McGraw Hill's Building Cisco Multilayer Switched Networks, designs and leads the execution of the test suites. He can be reached at john_bass@ncsu.edu. Sangram Kadam and Khurram Khan assisted with the testing.
Bass is also a member of the Network World Global Test Alliance, a cooperative of the premier reviewers in the network industry, each bringing to bear years of practical experience on every review. For more Test Alliance information, including what it takes to become a member, go to www.nwfusion.com/alliance.
Cisco's Catalyst 3550 midrange switch
Layer 3 switch delivers flawless wirespeed performance.
