Everyone has heard of "box fatigue," the concern of data center managers with too many single-function devices proliferating in the data center. Unfortunately, VCs do not get tired quickly, so there seems little hope of relief from the "box startup fatigue" suffered by our industry. And, while there is no doubt that appliance computing is a positive force for the next-generation data center, proliferation of these appliance startups is a danger to all of us vendors and end-users alike.
First, some of these startups' products offer little value, and are simply companies that shouldnt exist. This isn't a polite thing to say, but every reader can think of at least one such company, with a product that is really a minor feature. Another frequent case is a commodity product that offers only marginal price-performance advantage over Cisco Systems or Juniper/NetScreen. In a more efficient market, they would either be closed down or never get funded in the first place.
Second, there are simply too many companies doing more or less the same thing. Even if they have good technology and sound management, the resulting market fragmentation results in semi-starved, staggering startups barely capable of delivering and supporting competent products never mind pursuing ground-breaking innovations. Competing for customers consists of grueling bakeoffs of 5 or 10 vendors, followed by fierce price-cutting. In the end, with more startups competing for an ever-diminishing number of IPOs or M&A exits, the employee and VC shareholders have a proportionately lower chance of getting a return for all the hard work.
Third, many software vendors have recently tried to transform themselves into "data center appliance vendors" by installing their same software product on an off-the-shelf server, and shipping it to customers. Appliances are supposed to reduce complexity of installation and ongoing operation of certain functions, but that requires a different engineering design and discipline not merely shipping servers instead of CD ROMs. These "server appliances" confuse the market further and provoke end-user mistrust of all appliance vendors. As a result, most data center managers now assume that any new appliance is a pig-in-a-poke, simply a general-purpose server running Linux, Apache, MySQL, and the PHP script language (the so-called "LAMP" stack).
These and other factors make true innovation more difficult and stunt the development of the new data center. Companies designing custom silicon or clever network kernel algorithms are drowned in a sea of "me too" appliances, while end-users have to pick through 10 vendors in every new niche. While the natural, slow-acting forces of consolidation and attrition provide some population control, the venture and customer over-investment continues to add more data center appliance startups. Mysteriously, some of the companies funded sound exactly like ones that failed the week before. What, then, is the solution?
