CIOs already had their hands full dealing with modernizing their infrastructures and wrapping new services and value around the public cloud–juggling two speeds of IT–before serverless came along. Now, serverless computing has introduced a third mode of IT that needs to be addressed.
However, each of these three modes has its place in the enterprise. Let’s examine the three modes of IT operations and where executives should focus for the biggest return.
SEE: Quick glossary: DevOps (Tech Pro Research)
Mode 1: Legacy DevOps
This mode of operations is the portion of enterprise IT that Oracle, Exchange, and other monolithic applications run. In the pets vs. cattle descriptor, these applications are the pets. For most organizations, this model runs atop of a virtualization platform such as VMware vSphere, Hyper-V, or KVM.
Vendors such as HPE with SimpliVity, Nutanix, and Datrium look to optimize cost and operations for the related infrastructure. This mode is the least agile of the three modes. For this reason, CIOs look to shrink this portion of the operating budget. However, due to cost, risk, and technical complexity, many of these applications will remain for years, if not decades, to come. To shrink the operational cost of this mode, enterprise infrastructure groups look to automation to reduce the overhead associated with these applications.
I call this mode Legacy DevOps, as IT staff may find more value here than end users or developers. As Legacy DevOps improves, the cost of managing legacy applications recedes. CIOs may then shift the resources to Mode 2 and Mode 3 operations.
Mode 2: Cloud-native DevOps
Cloud-native is a broad category of infrastructure. These solutions include VMware Photon, Docker, Kubernetes, OpenStack, and Microsoft AzureStack. Cloud-native applications include both monolithic and micro-service application architectures, but they all share a common theme: The integration of the infrastructure and the application.
I call this the Netflix model of DevOps. In the Netflix model of operations, the application controls and monitors the infrastructure, and it also auto-scales the infrastructure as demand increases. Performance monitoring is built into the application. Over the years, vendors have started to provide turnkey solutions to provide similar capability. Cisco acquired Appdynamics, which focuses on delivering performance monitoring for cloud-native applications, and all the top-tier cloud providers offer an auto-scaling API to scale for demand.
Operations and development teams should work hard to mature this model as it offers both agility and flexibility from a provider perspective.
Mode 3: Serverless NoOps
I recently interviewed Nick Rockwell, CTO of The New York Times. According to Rockwell, the Times is making a big bet on serverless. It’s important, however, to differentiate between serverless computing and Functions as a Service (FaaS).
FaaS is a subset of the serverless delivery model. According to Rockwell, FaaS isn’t enough to build a complex application such as the NY Times Crossword app. Rockwell selected Google’s App Engine to build the crossword app instead.
Serverless has its challenges. Removing the burden of the underlining infrastructure also means removing the crutches of the underlining infrastructure. Engineers can’t simply add more disks or more CPUs to tweak performance. Just as applications will not seamlessly move from Mode 1 to Mode 2, Mode 2 applications will not seamlessly move to Mode 3. While serverless brings the most agility to application development, CIOs will need to consider how to spread resources across these very different modes of operations.