Abstract

3Com Corp. commissioned Tolly to provide independent validation of the NETBuilder II bridge/router's WAN traffic prioritization features. 

Tests validated the effectiveness of the prioritization feature when running various combinations of SNA, NetBIOS, IP, and IPX traffic across a 56 Kbit/s WAN link.

This Tolly Group report evaluates the WAN traffic prioritization capabilities of the 3Com NETBuilder II in mixed-protocol enterprise environments. Commissioned by 3Com and tested in March and April 1994, the study examined an experimental 7.0 software release running on paired NETBuilder II bridge/routers across a simulated 56Kbit/s WAN link. The objective was to validate whether the platform could guarantee predictable bandwidth for latency-sensitive SNA traffic while it competed with IP, IPX, and NetBIOS traffic on the same WAN connection. 

The results showed that the prioritization feature successfully controlled bandwidth allocation within acceptable error margins. In the first test, using source-route bridged (SRB) SNA plus routed IP and IPX traffic, the router was configured for a 60/20/20 split. Measured results, shown in the chart on page 1, delivered 71.1% of bandwidth to SNA, 12.3% to IP, and 16.4% to IPX, with total PPP WAN utilization of 55.94Kbit/s. In the second test, which used LLC2-tunneled SNA with IP and IPX at the same 60/20/20 target ratio, the usable bandwidth was affected by TCP/IP encapsulation overhead, but the allocation remained close to target: SNA received 54.6%, IP 19.9%, and IPX 21.7%, with total utilization of 53.9Kbit/s. In the third test, shown on page 2, SRB SNA and SRB NetBIOS were prioritized at 75/25, and measured results were 72.4% for SNA and 27.0% for NetBIOS, with total link utilization of 55.6Kbit/s. 

Tolly’s methodology used total data frames of 1,059 bytes for all protocols, with the SNA session corresponding to a 1Kbyte Request/Response Unit plus headers. IP and IPX traffic was generated with a Wandel & Goltermann DA-30 analyzer, while SNA traffic was produced by two OS/2 2.1 SNA/APPC systems running Communication Manager/2 and Bader Benchmark APPC applications. A Network General Expert Sniffer on the receiving LAN captured 30-second traffic samples to calculate protocol distribution and throughput. The test bed, illustrated on page 4, linked two 16Mbit/s token-ring LANs through two NETBuilder II routers and a simulated 56Kbit/s WAN. 

Overall, the report concludes that the 3Com NETBuilder II could enforce practical WAN bandwidth guarantees for SNA and other protocols in a multi-protocol internetwork, making it well suited for environments where mission-critical SNA traffic had to coexist predictably with general routed and bridged traffic.