Abstract

SMC commissioned Tolly to evaluate the EliteSwitch ES/1, with the main focus on demonstrating the performance benefits of a Token Ring switched backbone compared with a traditional bridged backbone. The project quantified aggregate client-server throughput in multi-ring campus-style Token Ring environments and examined whether a switching approach could relieve backbone congestion without requiring changes to end-user wiring, adapters, or communications software.  

The March 1995 report positions the EliteSwitch ES/1 as an alternative to expensive backbone upgrades such as 100Mbit/s FDDI. Tolly notes that replacing a traditional bridged backbone with a Token Ring switch can occur in the wiring closet while leaving desktop connections unchanged. The tested EliteSwitch ES/1 was a multi-port Token Ring switch configured for source-route bridging and designed to interconnect six 16Mbit/s Token Ring segments through an internal backplane. The product-specification panel on page 3 lists software version 2.5 S14 Jan. 11, 1995, support for up to 20 fully interconnected interfaces per chassis, the Texas Instruments TMS380 Token Ring chipset, 4 or 16Mbit/s Token Ring over STP and UTP, and additional support for Ethernet, Token Ring, FDDI, and HSSI. Supported bridging modes included Transparent Bridging, Source Route Bridging, and Source Route Transparent Bridging, while routing support included TCP/IP and IPX.  

Tolly’s methodology used Novell Perform3 version 1.61 invoked from nine clients connected to three NetWare 4.01 servers. Tests were run at frame sizes of 68, 512, 1,024, 2,048, 4,096, and approximately 4K burst frames using Novell’s PBURST function. Aggregate throughput was calculated by summing the throughput from each ring of clients to the servers. Traffic was measured with a Siemens K1100 Protocol Tester, and frame validity was verified with a Network General Expert Sniffer. The traditional backbone test bed consisted of four 16Mbit/s Token Ring segments linked by three Acsys TokenSwitch 3000 source-route bridges, while the switched backbone configuration used six 16Mbit/s Token Ring segments interconnected by the EliteSwitch ES/1.  

The results showed a substantial performance advantage for the switched design. Tolly states on page 2 that replacing the traditional backbone with the EliteSwitch ES/1 and dedicated server rings increased aggregate throughput by nearly a factor of three. The chart on page 4 shows throughput for the EliteSwitch ES/1 versus the Acsys bridge baseline: 3.95 versus 4.70Mbit/s at 68-byte frames, 11.81 versus 13.22Mbit/s at 512 bytes, 16.63 versus 14.91Mbit/s at 1,024 bytes, 23.07 versus 15.42Mbit/s at 2,048 bytes, 27.05 versus 15.54Mbit/s at 4,096 bytes, and 44.68 versus 15.57Mbit/s with 4K PBURST. Tolly highlights that the traditional backbone saturated just under 16Mbit/s, while the EliteSwitch ES/1 delivered up to 44.68Mbit/s, roughly equivalent to nearly three 16Mbit/s Token Ring rings operating in parallel. Overall, the report presents the SMC EliteSwitch ES/1 as an effective way to revitalize congested Token Ring backbones by dramatically increasing aggregate throughput without forcing wholesale network upgrades.