How we did it
 
|
|||
 | |||
|
|
|
|||
| |||
|
Also: |
We ran baseline tests in configurations with eight, 32 and 64 interfaces per chassis. These tests involved fully meshed traffic, meaning all ports exchanged packets. We ran each test twice using two packet sizes: 64 and 1,518 bytes. Spirent created custom scripts that offered up to 224 unique IP source addresses to each switch interface. In the 64-port full-mesh tests, there were more than 14,000 addresses involved. In the baseline tests we measured throughput, latency, jitter and packet sequencing.
Trunking
We asked vendors to link two chassis with one or more aggregated links. We connected the Smartbits to 32 interfaces on each chassis and offered traffic across the trunks. We began with a single trunk consisting of eight ports and reran the same tests with two trunks of eight ports apiece. The Smartbits represented 224 unique IP addresses per port. With 64 ports total on the test bed, there was traffic from 14,336 unique hosts involved. To examine how these switches handled the addition or loss of a port from an active trunk, we defined a seven-port trunk between two chassis. From each of 64 Smartbits card, we offered exactly enough traffic to saturate the seven-port trunk.To test the addition case, we enabled an eighth port in the trunk 10 seconds into the 30-second test duration. In the subtraction case, we began the test with eight ports in the trunk and removed one of these 10 seconds into the test. We repeated the three steps (baseline, add and subtract) with short and long packets.
Failover
We offered traffic at a rate of 1 million packet/sec in each direction across a switch link for a duration of 60 seconds. Ten seconds into the test, we removed the primary circuit, forcing Open Shortest Path First to reroute packets over the back-up link. Determining cutover time was a matter of simple math: At rate of 1 million packet/ sec, the switches would drop one packet for each microsecond between the failure of the primary link and the establishment of the secondary link. We ran the failover test in two configurations: once with a single link between chassis, and again with an eight-port trunk between the boxes.
Quality of service
We set up a test involving three classes of traffic and two chassis connected over a congested link. We offered the three classes of traffic in one ratio (1-to-10-to-10 for the high-, medium- and low-priority flows, respectively), and asked vendors to deliver the classes of traffic in a different ratio (2-to-12-to-7). We also asked vendors to deliver the relatively small amount of high-priority traffic with no packet loss. Then we ran the test a second time, this time with no high-priority traffic. This time we offered equal amounts of medium- and low-priority packets, and asked vendors to deliver the traffic in a 5-to-3 ratio. Vendors had to use the same configurations in both test runs. To create congestion, we offered enough traffic to 21 interfaces on each chassis to present a 2-to-1 overload on the link between the boxes. We repeated the quality-of-service tests with a single Gigabit Ethernet link and a seven-port trunk between the two chassis. In addition to preserving some packets and junking others, we expected the switch/routers to mark the different packet headers' IP precedence fields to indicate the relative priorities of each traffic class.The complete test methodology is available by clicking here. Network World acknowledges the support of test equipment vendors who supplied key infrastructure and support for this project. They include Spirent Communications of Calabasas, Calif., which supplied its Smartbits 6000B chassis and 3201B Smartmetrics cards. Also, Brooks Hickman, a member of Spirent's technical staff, developed custom traffic scripts to run on the Smartbits. Network Associates of Santa Clara, supplied its Sniffer Pro 4.6 protocol analyzer for use in traffic monitoring. Fluke, of Everett, Wash., supplied its Fluke One-Touch cable tester, which we used in debugging test bed infrastructure.
RELATED LINKS
Newman is president of Network Test in Hoboken, N.J., an independent benchmarking and network design consultancy. Network Test's clients include large firms, service providers and publications; to preserve its independence the company does no vendor testing. Newman is active in the IETF's benchmarking working group, and is co-author of a draft specification on network-layer quality-of-service mechanisms. He can be reached at dnewman@networktest.com.
Newman 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.
The trouble with trunking
Our tests show that high-end switch/routers run lightning fast - until you set up trunking.
Interactive scorecard and NetResults: Directory administration tools
Find out which product would best suit your needs.
