Can containers threaten VMware’s beachhead? Yes. Combined with an exploding ecosystem, they address the top use-cases and add something additional.
I started working with VMware during its first ESX beta (nearly 15 years ago!). My first start-up, co-founded with Dave McCrory, created multi-host automation for VMs. This ultimately came to be known as Cloud. Back then, virtualization was considered risky, slow and insecure. One of the biggest adoption hurdles was that no one agreed on the “right way” to run VMware.
Even in the wild, minimal first days, VMware was revolutionary
VMware changed our life in operations. It smoothed out all those differences between systems and allowed users to compress five servers into one. Suddenly, our server images were portable and super dense; however, operational improvements were not enough to make ESX compelling.
Why? ESX was difficult to configure and manage with minimal APIs. We spent days and months trying to design inexpensive yet reliable reference architectures only to find they did not work for our customers. Even tracking the VMs was hard because there was no multi-host API.
That changed overnight with live migration, VMware’s killer feature. vMotion allowed you to isolate your workload from the physical server and keep it running (or get near-instant recovery) if there was a failure by moving your workload seamlessly between different physical hosts. There was a cost: it required a shared storage configuration that was prescriptive enough to kill variation: you needed a SAN and 8-16 diskless servers (aka blades) per pod.
Consistent Ops, Density and vMotion turned VMware into an IT juggernaut.
Can containers threaten this beachhead? Yes, if they address all three use-cases and add something additional. By themselves, containers would not be not a threat; however, in combination with other changes, I believe they are.
Here are four threats that I see creating a disruption for VMware:
- vMotion threat > the new host of container orchestration tools (building the list is left to the reader) offers to create durable applications using automation. This requires retooling and rewriting but the ROI of portable elastic applications is very high.
- Consistent ops threat > the rising ability to automate directly to metal provides operational value. With metal ops, we’ve reduced the need for virtualization to homogenize the environment.
- Density threat > containers are lighter weight than VMs by many times. This translates into powerful performance and over-subscription ROIs.
- New threat: developer velocity > unlike VMs, containers are so lightweight that developers can run them directly on their laptop. This creates a potent “keyboard to production” cycle time benefit.
Considering the cost and overhead added by virtualization, I believe IT groups will be taking a very serious look at containers (or related container-like hyperVMs) as the keystone to a powerful alternative.