Sun and Their Servers

I was reminded of the time in my career when I found myself making server technology choices when I came across this article today. It's been a long time since the glory days for Sun Microsystems, and about as long since I was choosing servers. Much has changed since then.

Sun built an enviable reputation for server technology at a time when the market for servers was growing explosively. It was driven by the web boom and also by a shift in the telecom equipment sector from custom processors to standard processors and then complete hardware systems. Sun's servers were of high quality for the intended applications, used an operating system (Solaris) that was Unix based but extended to handle real-time and high-volume transaction processing. Like IBM and other industry leaders, Sun invested heavily in technology research and for their pains has ample intellectual property (patents) on high-availability software and hardware methods which are so important to commerce and telecommunications.

Then the industry changed. As the telecom business became more competitive and with the low cost of entry to participation in web services, there was a drive to lower equipment costs. Sun was slow to respond. This created an opening for Linux (OS software) and Intel, Dell and all the rest of the component and system players. Yet Sun held the technology advantage for quite some time. Linux, for example, was not well-suited to real-time applications in the early days, though that has long since changed.

While costs were coming out of their competitors' products (and quality was improving), Sun was held back by their custom hardware and software; because they were the only users of it, they did not achieve the volumes and thus lower-costs of the competition. In time they did change, though by then competition was full blown. It would have happened anyway though it is arguable that shifting strategy earlier would have left Sun in a more dominant position.

An important factor that further harmed them was that they misread the changing market. The focus shifted from high-reliability servers to racks full of less reliable but very cheap servers. While Sun built more and more expensive and fully-redundant servers in their Netra line, culminating in the ft-1800, customers were opting for alternatives. Cheap servers could be discarded when they failed and distributed over multiple networks and locations for superior service availability. Google has perhaps taken this to an extreme with their buildings full of ultra-cheap servers, and the transaction distribution technology to make it all work. Those expensive Netra servers did not fit the changed requirements.

I made a similar decision at one point to switch product strategy from internally redundant servers to redundancy over multiple servers. My decision was driven by customer feedback. They all loved Sun but laughed at the price tag of the ft-1800 and similar technology. These servers were nevertheless an impressive engineering achievement.

Now, as the author of the referenced article states, Sun appears to be drifting into irrelevance. What a shame.