WAN scalability: More than just bandwidth

At first glance, the scalability of the WAN seems to be pretty simple – a T-3 circuit has a lot more capacity than a T-1 circuit, and an OC-3 circuit has yet more capacity. So what’s the issue? The issue is that WAN scalability is impacted by many factors other than just the ability to increase the bandwidth of the WAN circuits. To make matters worse, some of these factors are very subtle.

The issue of WAN scalability is not new. In the early 1980s, enterprises started to make use of X.25 networks. In many cases, these networks were comprised of 9.6Kbps links to the branch offices, and one or more 56Kbps links into the data center. While these link speeds seem infinitesimal by today’s standards, it is important to realize that at that time few people had access to e-mail, the Internet was used only by the research community, the chatty protocols that are now common (i.e., HTTP) did not exist and nobody was transmitting 10MB PowerPoint files.

Companies that used X.25 networks soon found out that link speed was only one factor that impacted the scalability of the WAN. For example, companies whose employees came to work and logged onto the network first thing in the morning and stayed on all day soon often ran into scalability issues. In particular, early X.25 networks had a relatively low limit on how many simultaneous virtual circuits could be supported. Because of this low limit, many companies experienced the situation where the WAN could not scale to accept another user because all of the virtual circuits had been assigned. This was particularly counter-intuitive because in many cases the users who were logged on were not transmitting much if any data. Hence, the transmission links were not heavily utilized, but the network had reached exhaust.

The number of supportable virtual circuits was not the only subtle factor that impacted the scalability of X.25 networks. Another factor was that most X.25 networks could only support a relatively small number of call set-ups per second. As a result, many companies that were doing a lot of transaction processing found themselves in the situation where their transmission links were lightly loaded, but the network could not support all of the possible business transactions because the network had reached its limit of call set-ups per second.

In the next WAN newsletter, we will discuss scalability in a more contemporary setting – the scalability of network and application acceleration solutions.