CloudPhysics has announced a set of predictions for 2016 based on findings from its Global Data Set. The CloudPhysics Global Data Set contains more than 100 trillion samples of configuration, performance, failure, and event data from companies of all sizes and industries. This anonymized Global Data Set provides collective intelligence and unprecedented insights into user behavior, and the structure and operations of virtualized data centers around the world.

"The CloudPhysics Global Data Set provides tremendous insights into the range of technologies deployed across today's data centers, as well as how they are functioning in practice to support today's enterprise workloads," said Chris Schin, vice president of product management at CloudPhysics. "As we analyze the Global Data Set, we can make data-driven predictions about what will happen in 2016. These are not mere guesses, these are based on observed tendencies across a large set of actual, anonymized data centers."

Prediction 1: Enterprise cloud migrations will hit an inflection point in 2016, as enterprises use analytics and big data to right-size their on-premises VMware virtual machine (VM) configurations.

Among virtual data center administrators, it is well known that VMs are often configured conservatively, meaning that they are configured with far greater resources than they are expected to actually consume. This situation is management for on-premises virtual environments because administrators can "load up" a host (or a data store) with any number of over-resourced VMs, and resource consumption can be tracked and managed at the aggregated, host level.

However, when virtual administrators are asked to consider migrating these VMs to the cloud, all too frequently they use those, sometimes massively, over-resourced VMs to "size" their cloud equivalents. This can vastly overstate the OpEx costs of moving on-premises workloads to the cloud.

Here are some data points uncovered from the CloudPhysics Global Data Set, when comparing actual resource utilization against resource configuration:

When looking at resource utilization:

  • On average, 98.1% of VMs used less than half of their configured CPU (and 95.4% used less than 25%)
  • At peak*, 89.3% of VMs used less than half of their configured CPU
  • On average, 99.4% of VMs used less than half of their configured memory (and 94.4% used less than 25%)
  • At peak, 92.0% of VMs used less than half of their configured memory

Using analytics to right size VMs translates into meaningful dollar savings. In a simple, CPU-centric example — suppose an organization deploys 50 VMs locally, each with an arbitrary 16 CPUs each, but suppose that those 50 VMs only utilize four CPUs each. Using AWS' "compute optimized", on demand pricing, the savings that organization would realize by right-sizing those VMs when migrating to the cloud would be over $500,000 per year.

Once Enterprise IT leaders rigorously examine actual resource utilization and use data to "right-size" their workloads, the costs of the cloud will come into clearer focus, and migration of workloads to the cloud will increase dramatically.

Prediction 2: VMware ESXi 6.0 will reach 50% market share quicker than ESXi 5.5 did.

It can be argued that the rate of release of "Critical" and "High Importance" Knowledge Base Articles can be viewed as an indication of a product's stability — the more frequently these types of KB articles are published for a given product, the lower the implied stability of that product since critical KB articles are typically urgent calls-to-action associated to issues discovered. As a result, one would expect to find a higher rate of these types of KB articles within the first months of a new release of a technology, followed by a levelling off as the product stabilizes.

This is precisely what was seen in the weeks following the ESXi 5.5 and ESXi 6.0 releases, but the stabilization process occurred far more rapidly following 6.0 than 5.5, implying that ESXi 6.0 was a more stable release when launched. For example, it took over two months for 5.5 to reach a level of < 10 KB articles per week, while it took 6.0 a mere two weeks to reach that same level of implied stability.

The more stable a release is, the more rapidly it is adopted by the market. Therefore, we predict that VMware ESXi 6.0 will achieve market dominance more quickly than 5.5 did.

Prediction 3: VMware ESXi 6.0 will become the dominant ESXi platform deployed globally by August 2016.

Based on our analysis of adoption rates of ESXi 5.1 and 5.5, a new release of ESXi achieves dominance in global data centers approximately 18 months after its initial deployment into the market. This suggests that ESXi 6.0 will be the dominant version of ESXi deployed globally in September of 2016, given its March 2015 release date. However, given the rapid stabilization of ESXi 6.0 (described in Prediction 2) and the additional discovery that ESXi 6.0 reached 10% market share more quickly than ESXi 5.5 did, we believe that ESXi 6.0 will achieve 50% market share more rapidly than 18 months — we predict it to achieve this milestone in August of 2016, but it could occur even sooner.

* metered at 99th percentile

 

This article was originally posted “CloudPhysics Unveils 2016 Predictions” from Cloud Strategy Magazine.