v. 1.8 Copyright © 1999 CMP Inc. Vendors are encouraged to comment on this document and any other aspect of test methodology. However, Data Communications reserves the right to change the parameters of this test at any time.
By Bob Mandeville, ENL, and David Newman, Data Communications
This evaluation of high-performance routers will compare new products aimed at accelerating IP routing on service provider networks. The tests described here include baseline measurements of router forwarding rates, packet-by-packet latency, and jitter; routing table lookup times; performance during BGP4 flapping conditions; and failover of OSPF routing.
All devices in this test will use POS (packet over Sonet) interfaces operating at OC-12c and OC-48 rates. An ancillary goal of this test is to compare OC-12 POS performance with that of ATM OC-12 interfaces from earlier tests.
As shown in the following diagram, each vendor will supply four routers, each equipped with 12 OC-12 (622-Mbit/s) interfaces, three OC-48 (2.4-Gbit/s) interfaces, and fast Ethernet interfaces to inject routes. Vendors may also supply trunked OC-12 interfaces in lieu of OC-48 interfaces. All OC-XX interfaces must support packet over Sonet (POS) framing using PPP (point-to-point) protocol encapsulation.
IP packets will be offered to each device using a Smartbits 6000 traffic analyzer from Netcom Systems Inc. (Chatsworth, Calif.). BGP4 and OSPF router table entries will be offered using a QA Robot system from QOSnetics Inc. (Portsmouth, N.H.).
This series of measurements will determine the forwarding capacity, latency, and jitter characteristics of the routers under test.
Router interfaces will be configured with IP addresses 192.1.XXX.254, where XXX is the number of the port. Router A will use ports 1-12; Router D will use ports 13-24; Router B will use ports 25-36; and router C will use ports 37-48. The Netcom Systems Smartbits interfaces will use addresses of 192.1.XXX.253.
The routing table of each device under test will be loaded with 36,864 prefixes, with each prefix containing one route. Then bidirectional partially meshed traffic will be offered to all 48 OC-12 ports of the test bed. Load will be applied at full wire rate, which in turn will fully load the six OC-48 backbone links that interconnect the routers. Traffic will be generated in 60-, 340-, and 1,504-byte IP packets. The 60- and 1,504-byte packet lengths represent the minimum and maximum Ethernet frame sizes, as transmitted on Sonet; the 340-byte frame length represents the mean packet length of the top 10 TCP and top 10 UDP applications on an Internet backbone link, as reported by CAIDA.
Forwarding rates will determined for longest match lookups with /16 and /22 prefix lengths. These tests will be repeated with a routing table holding 73,728 prefixes, again with one route per prefix. To discourage the possibility of route aggregation, the QA Robot will inject noncontiguous prefixes, and verify that the devices under test advertise the same number of prefixes as those inject by QA Robot.
The following measurements will be made on all outbound OC-12 interfaces:
1. Per port forwarding rate for all 48 OC-12 ports under test, using 50-, 340-, and 1,504-byte packets and /16 and /22 prefix lengths.
2. Minimum, maximum, and average latency measured on a packet-by-packet basis on all ports.
3. Jitter (latency variation) will be calculated using results from the latency test.
This test will determine the maximum number of BGP4 routes that the devices under test will support.
All devices will be configured to run BGP4 . The QOSnetics QA Robot tester will be attached to one interface of routers A and B. The routing tables of the devices under test will be cleared. Then 40,000 /22 prefixes, with one route per prefix, will be injected by the tester to Router A. The correct learning of the routes will be verified on the BGP4 instance running on the QA Robot unit attached to Router B. The number of prefixes/routes to be injected will range from 40,000 to 520,000 in increments of 40,000.
The test will report the maximum number of routes learned by the router under test.
This test will determine the effect, if any, on forwarding rate of the devices under test in face of network instability--i.e., route flapping.
All devices under test will be configured to run BGP4. The QOSnetics QA Robot tester will be attached to one interface of routers A and B to generate route flapping and to verify table entries at every flap event. This test will be run with 48 peering sessions per router, and rerun with 96 peering sessions per router.
During these tests, the Netcom Systems Smartbits 6000 will offer 64-byte packets to all POS interfaces in a partially meshed pattern at line rate. Then this procedure will be followed:
Step 1. Inject 73,728 /22 prefixes into the routing table of Router A. Each prefix has three routes--a primary, secondary, and tertiary routes.
Step 2. Withdraw 12,288 primary routes from the routing table.
Step 3. Verify that traffic is routed over 12,288 secondary routes.
Step 4. Verify that traffic is routed on all other available routes.
Step 5. Re-inject the original 12,288 primary routes after an interval of 30 seconds.
Step 6. Verify that all traffic is routed over the primary paths.
Step 7. Remove and re-inject 12,288 routes from the routing table continuously in 30-second intervals for a period of 120 seconds. The routes removed in each iteration will be noncontiguous entries in the routing table.
The test will report the maximum forwarding rate the devices under test can achieve over stable and unstable paths during flap events.