What are your IPX-pectations?

By Kevin Tolly
Network World, 04/01/02

IPX routing: relevant or relegated to the junk heap of networking? Some recent events have brought forth this question.

From the beginning of "LAN time" through the late 1990s, the importance of IPX rose in conjunction with the popularity of Novell's NetWare server operating system. As NetWare became ubiquitous, so did IPX. Indeed, the fact that Novell's "native" protocol was routable was a strong selling point against Microsoft, which had built its server on the unroutable NetBEUI, broadcast-based protocol.

At some point in the late 1980s or early 1990s, IPX no doubt became the second-most prevalent protocol in enterprise networks, after IBM's SNA.

But as Novell and Microsoft fought for server supremacy, the rise of the Internet - and IP - neutralized Novell's routing advantage. Microsoft and Novell had to react to this new reality and overlay their application services onto an IP transport.

As the Layer 3 switch era dawned, IPX rapidly was falling in importance for two reasons. First, Microsoft was in the process of dethroning Novell as the king of servers. Second, Novell had developed a version of NetWare that delivered its services over IP rather than IPX. The "pure IP" version of NetWare, Version 5, shipped Sept. 14, 1998.

While customers did not convert overnight - and some didn't convert at all - the clear direction of an IP-everywhere corporate environment anchored to the Internet no doubt put IP conversion high on the lists of loyal NetWare customers.

IPX was in its twilight. Not long after the shipment of NetWare 5, The Tolly Group floated the idea of a multivendor benchmark of IPX routing performance. The response from the industry was united. Even though they all claimed to offer IPX routing support, no one was interested in participating in such a benchmark.

While we constantly benchmark Layer 2 and Layer 3 IP performance for switch vendors, not one requests that we run IPX numbers. Case closed, or so I thought.

In recent weeks, the subject has come up again in conjunction with a request for proposal qualification project on which The Tolly Group is working. Among the requirements is qualifying the packet throughput of IPX routing on Fast Ethernet and Gigabit Ethernet workgroup and core switches.

While doable, one has to ask: "Why?" NetWare 5.1 and 6 have since followed the first pure IP release so one has to believe that NetWare and IP work smoothly together.

Given the nonexistent demand for IPX throughput testing, we had no canned scripts ready. And, when we hunted through the test-tool documentation, we found that we had to craft all the scripts ourselves from scratch. The hidden message there is that so few people are asking for IPX scripts that the test-tool vendor has decided not to spend the time developing them.

So are you "high-end" IPX users out there? If so, what are you doing with IPX that could not be done better with IP? Should vendors be encouraged to prove their IPX routing throughput capabilities as well as their Layer 2 and IP, or are you looking for low-end routing performance to support a dwindling number of legacy NetWare/IPX applications? Speak up or forever hold your peace.