Abstract

Cisco Systems, Inc. commissioned The Tolly Group to benchmark the throughput and latency of its Catalyst 1600 Token Ring switch versus that of the Bay Networks' SpeedSwitch 100.

The testing demonstrated conclusively that the Catalyst is capable of very high throughput and very low latency. The Cisco Catalyst forwarded more than 42,000 frames per second per port when source route bridging four streams. Tests of 16Mbps Token Ring were run with 2, 4, and 8 ports of traffic. 

Summary:

This September 1995 Tolly Group report compares Cisco’s Catalyst 1600 Token Ring switch to the Bay/Centillion SpeedSwitch 100 under source-route bridging at 16 Mbit/s. Catalyst 1600 delivered markedly higher throughput at small frames and matched wire-speed at large frames, while also exhibiting dramatically lower latency due to cut-through switching vs the competitor’s store-and-forward design. In quad-stream 8-port tests, Catalyst reached 171,600 fps at 28-byte frames (38.44 Mbit/s) vs 48,000 fps for SpeedSwitch (10.75 Mbit/s), and 100,000 fps at 64-byte frames (51.20 Mbit/s) vs 44,000 fps (22.53 Mbit/s). At larger frames both achieved wire speed: 1,024-byte aggregate 62.91 vs 61.93 Mbit/s and 3,972-byte aggregate 63.55 vs 63.55 Mbit/s. In single-stream tests, Catalyst posted 45,150 fps at 28-byte frames (10.11 Mbit/s) vs 26,000 fps (5.82 Mbit/s), and 25,000 fps at 64-byte frames (12.80 Mbit/s) vs 19,500 fps (9.98 Mbit/s).

Latency:

Catalyst’s cut-through architecture slashed latency by up to ~98% at large frames: ~48 µs vs ~3,303 µs at 3,972 bytes and ~46 µs vs ~1,002 µs at 1,024 bytes. Even at small frames Catalyst maintained lower delay: ~44 µs vs ~229 µs at 28 bytes and ~48 µs vs ~265 µs at 64 bytes.

Methods & setup:

Tests used up to eight 16 Mbit/s Token Ring LANs with unicast IPX traffic, measuring two scenarios: single-stream (2 ports) and quad-stream (8 ports). Traffic generation employed Wandel & Goltermann DominoLANs and LANQuest FrameThrowers; zero-loss throughput (1–2% margin) and per-port/aggregate FPS and Mbit/s were reported. Latency was measured with Azure LANPharaoh across ingress/egress rings. Results highlight Catalyst 1600’s advantage at small frames and consistently lower latency, with both solutions reaching wire speed at larger frames.