Will virtualization, multicore, and cloud computing trends send x86 architecture server and processor volumes down for the next decade? It certainly is a realistic scenario – and perhaps the most likely.
At Gartner, we spend a lot of time trying to understand future scenarios, the likelihood of each, indicators that a scenario is likely to occur, impacts on our clients, and what our clients should do. We’ve studied the impact of virtualization on the server market since virtualization was first introduced <begin chest-thumping>and Gartner was the first firm to point out the negative ramifications of virtualization on server volumes<end chest-thumping>. But we’re getting to the moment of truth.
With the exception of the economic collapse in 2009, server volumes have been dependably growing for years. However, virtualization rates are hitting a point that the negative effect of virtualization on the server market are becoming unmistakable. Not in five years. Now.
2010 was a good year for servers – nearly 9 million were sold. My contention is that if virtualization didn’t exist, there would have been 13, or 14, or 15 million sold.
The engine of server market growth has been the growth of workloads. Since 2004, the compound annual growth rate (CAGR) in workloads has been about 16 percent. 2010 was certainly a much better year than that – but if you factor in the the volume decline in 2009, the growth in 2010 just exactly made up the difference.
If the workload CAGR remains steady, server volumes will start to decline in 2011, and we won’t see 2010’s volumes again in this decade.
The good thing – virtualization (and cloud computing) makes it easier and faster to deploy a workload, and that has a tendency to increase the workload CAGR. However, even accounting for faster workload growth, 2010 is either at or near the peak of server volumes for the next ten years.
However, if Moore’s Law is going to be driven by increasing amounts of cores, those cores are going to need VMs to leverage them. Multicore is going to drive higher virtualization densities, and even fewer servers.
What will it take to drive server volumes up? Low virtualization growth, high workload growth, low virtualization densities. A combination of factors that seems unlikely.
Bottom line – there are a number of realistic scenarios for server volumes in the next decade. Each scenario will drive different vendor behavior (and results), pricing, and end user strategies. But – anyone want to place a bet? I’m blogging it, so I’m placing mine right now.