3 Horsemen of the Network Apocalypse: Virtualization
Ours is an industry fueled by constant technical evolution. We like to frame it all as “innovation,” but not much of the current churn is innovative. CPUs get more capable, memory densities go up, transmission speeds increase. Every year we can buy a lot more technology horsepower with the same money. I call it predictable evolution. It’s the constant march that puts food on our tables and leaves mere mortals at our whim because we (at least kinda) understand the new capabilities of technology while they, our clients, generally don’t.
On top of that constant march, that steady IT evolution, once in a while there comes a new twist on technology that is truly innovative and actually disruptive to doing business as usual. When that happens, what we know goes out the window.
What Killed the Server Market?
The last really disruptive technology for IT was server virtualization. Turn your mental clock back a decade or so and think about the Internet bubble. Sun was busy “putting the dot in dot com.” HP and Intel were hedging their server architecture bets with Itanium — if only they could build a compiler to take advantage of the EPIC architecture (they couldn’t). IBM thought itself so far ahead of the game with its Power architecture that it talked both Apple and Microsoft into using IBM chips in some of their consumer products.
As Intel laid low the RISC/EPIC chips by bringing the performance and features of the x86 architecture on par, VMware negated most real operating system advantages offered by HP-UX, Solaris and AIX. No more need to tie OS and hardware tightly together and endure a 10x penalty to get reliability and performance.
Further, VMware exploited x86 performance by making workloads movable. Now, not only is the CPU a commodity, so is the entire box. If a server goes south, the software layer – VMware, Hyper-V and the rest – just moves the workload to another cheapo box.
With a little bit of architectural awareness at the time of its design, almost no application needs the highest performing and most reliable servers. As a result …
… a sweet spot has developed for workhorse servers at around $10,000. If the server vendors add too many reliability (or other) features and move their systems out of the sweet spot, the product doesn’t sell too well.
Suddenly those high flying hardware companies were forced to build high-power commodity boxes with comparatively tight margins. The result: IBM sold its x86 server line to Lenovo, HP is a shadow of its former self, Dell has sought the comforts of private ownership and Sun is a pathetic drag on Oracle earnings.
That’s disruption.
The folks at VMware, who I credit with upsetting the hardware apple cart with a lot of help from Intel, probably didn’t set out to ruin the server business for IBM, HP, Dell and Oracle. If anything, VMware had higher-level goals, like improving server usage and manageability, something that would change the relationship of the server vendors to their customers — but not necessarily strangle their cash flow.
Virtualization also paved the way for the second of the three horsemen of the netpocalypse: software-defined networking.
**Editor’s Note: Look for part two of Art’s series on Wednesday, Feb. 25.**
