Abstract

Cisco Systems commissioned The Tolly Group to evaluate router traffic prioritization on Cisco 4000 routers, with the main focus on independently validating new custom-queuing and bandwidth-management capabilities in multiprotocol WAN environments. The project examined how effectively Cisco’s software could guarantee predictable throughput for SNA traffic while simultaneously carrying other protocols such as NetBIOS, TCP/IP, IPX, and XNS across a constrained 56Kbit/s WAN link.  

The April 1994 report describes a series of five test suites run using Cisco 9.21 software on a pair of Cisco 4000 routers. The tests used approximately 1K frame sizes for all protocols and an SNA maximum LLC window size of eight. Traffic was generated across two 16Mbit/s Token Ring LANs connected through the routers and linked by a 56Kbit/s WAN connection. Two OS/2 SNA/APPC systems created the SNA traffic, while Wandel & Goltermann DA-30 analyzers generated IP, IPX, XNS, and NetBIOS traffic. Tolly increased the offered load until the routers began to drop frames and then measured the traffic crossing the WAN using a Network General Expert Sniffer.  

The report highlights that Cisco’s custom queuing successfully enforced predictable bandwidth shares under mixed-protocol load. In the SNA and NetBIOS test shown on page 1, bandwidth was configured at 70% for SNA and 30% for NetBIOS. Tolly measured 66.24% for SNA and 33.76% for NetBIOS, demonstrating close adherence to the configured values. Total throughput for the link was 50.2Kbit/s, slightly below the nominal 56Kbit/s rate because of TCP/IP encapsulation overhead across the WAN.  

Additional custom-queuing tests confirmed similar behavior across more protocols. In a four-protocol test using IP, IPX, SNA, and XNS with each configured for 25% of WAN bandwidth, Tolly measured results clustered tightly around that target: about 25.7%, 25.0%, 24.8%, and 24.6%. In a three-protocol test with SNA configured for 60% and IP and IPX for 20% each, measured throughput was about 50% for SNA and 25% each for IP and IPX. Although Tolly notes that this result was not as exact as the earlier trials, it still demonstrated that custom queuing was working and that protocol throughput without it would be unpredictable. Overall, the report presents Cisco 4000 custom queuing and SAP prioritization as an effective way to protect SNA traffic and deliver controlled multiprotocol bandwidth allocation across low-speed WAN links.