Stop Arguing About Cars and Start Managing Fleets

by Dan Olds

A lot of what I hear and read from IT analysts, vendors and the industry press reminds me of guys in my high school arguing about cars. Ford versus Chevy, superchargers versus turbochargers, and even a few posers arguing the merits of Ferrari versus Lamborghini engineering and performance. While these arguments were fun-filled and passionate, they were pretty much irrelevant--particularly since roughly 80 percent of my buddies are probably driving SUVs or have a minivan in their garage (the horror).

To me, many technology discussions today are equally irrelevant, at least to commercial customers who use technology as a business necessity. To use my auto analogy, business customers are like rental car firms--they need cars, but aren't all that interested in how fast they go or the cool technology that makes them work. What they need are solid, reliable cars at good prices from manufacturers who provide great support. While vendors and analysts concentrate on system speeds, feeds, and arcane design points, data centers (like our rental fleet manager) are increasingly focused on how IT products and vendors solve business problems by bringing higher availability, better return on investment, lower complexity and greater flexibility of IT operations.

The server market and, of course, servers themselves, have changed radically in the last ten years. Back in the mid to late 1990s, server marketing and customer decisions were all based on system performance, scalability, and size. There were considerable differences between what particular system types or vendors could and could not provide. Just like how auto makers back in the day could trumpet things--"Get off the crank! Ours starts with a key" or "Now, with two headlights!"--server vendors could market exclusive scalability and performance capabilities. Some systems scaled and others simply did not, and some processors were much faster than others. Performance and scalability were much more important back in those days since customers routinely had applications that needed to scale beyond the limits of the largest systems available. System availability was also a mixed bag; some systems could provide four "9s" with clustering and plenty of luck while others were fortunate to get to a single "9." (Those nines vendors talked about all the time referred to how many nines were in the uptime percentage figures--99.9 percent, 99.999 percent, and so forth.) There were lots and lots of server vendors to choose from, each bringing their own architecture and operating system to the party.

Over time, everything changed. Most systems today, particularly those based on Unix and OS/400, are big enough, scale well enough, and are fast enough to handle virtually any customer workload. In fact, these systems are so large now that customers are using them to host multiple workloads like mainframes. Availability is also much better; properly managed, all the Unix operating systems should be able to turn in 99.9 percent uptime with 99.99 percent or better with clustering.

We find ourselves living in a world of parity; at least as far as Unix, OS/400, and VMS servers are concerned. What I mean by this is that as far as base technology goes, we have a level playing field. The Windows and Linux operating systems have made some progress, but still have to make up some ground in terms of scalability, availability and workload management in order to compete head-to-head with the dominant Unixes--Solaris, AIX, and HP-UX--or IBM's OS/400-based iSeries platform, which might as well be Unix in terms of its uptime, resiliency, security, and sophistication. (To a more precise way of thinking, it is Unix that had aspired to be--and eventually became--like the VMS and OS/400 operating systems that predate Unix as commercial platforms. Unix may have been created in 1969, but it was not an enterprise-class platform until the late 1980s.)

If we were to go to IBM, Sun, and HP and ask them for a highly performing server with solid workload management, high availability, good scalability, etc., each vendor would be able to tell us that they can provide a box that meets all of our needs. Gone are the days when one vendor had a big technical advantage over everyone else or any product level competitive advantage that can be sustained. There are, of course, differences in how vendors apply their various technical features and capabilities. Some will argue that Sun Microsystems Solaris containers and dynamic domains are better than IBM's logical partitions or Hewlett-Packard's vPar virtual partitions and nPar hardware partitions. But much of this differentiation is based on customer experience and not on objective technical attributes. We're at the point where it's really difficult to prove that one technical feature or server is better than another. In a broad sense, current systems are fast enough, available enough, and manageable enough to satisfy the vast majority of customer needs.

So if this is true, what are the implications? In my mind, this means that the server market is going to become a whole lot more like the auto market. Sports cars will compete with sports cars; trucks will compete with trucks, etc. System vendors will find it much more difficult to successfully differentiate along technical lines--at least when selling to corporate customers (who usually move fairly slowly), since their competitors will probably be able to offer the similar technology within a quarter or two.

Using the car analogy, I believe that enterprises should, and will, act more and more like the rental car fleet managers discussed above than the teenage boys in shop class. In other words, while the boys are sitting outside of shop class arguing Ford versus Chevy or BMW versus Mercedes, the fleet manager is making his buying decisions based on how well the automaker helps the company solve their problems and build their business.

In IT sales, this will manifest itself in customers questioning how particular system vendors add value above and beyond simply supplying the hardware, operating system, and user manual.

In recent conversations, large Unix customers told me how they are looking to their system vendors for help in resolving a variety of operational challenges. Most of them cited goals such as raising anemic system utilization rates, lowering system management labor requirements and automating service level management. In my mind, this is the new battlefield in the system sales war. What applies to these large Unix shops applies equally well to an i5/OS shop, an OpenVMS shop, and a mainframe shop.

The vendor or vendors who help customers streamline their operations, get more value out of existing equipment, and ultimately lower IT costs will win the battles and increase their share of data center floor space. Those who continue to sell in the same old way--"this box is X percent faster than what you have, sign here, and press hard because the third copy is yours"--will find themselves on the losing end of the deals over time.

Dan Olds is the founder of Gabriel Consulting Group, an IT industry analyst firm that focuses primarily on IT as it relates to business, showing companies how technology can be deployed in a more efficient and effective manner. You can contact him through Gabriel Consulting's contact form.