Here are two predictions for 2016 around software development: (1) DevOps will make big inroads into the Fortune 1000 – but not much on the “Ops” side; and (2) DevOps will continue to drive new development and testing models, but it is not clear if it will influence production deployment and operations yet.
We at QualiSystems are seeing a real commitment to DevOps among our Global 500 customers. We expect to see an acceleration of this shift in 2016.
The shift to DevOps is putting developers in control over operations people. This is driving the tools toward more developer-friendly interfaces and away from the “simple” operations button. While this is great in engaging many developers, it might be hindering actual deployment into production.
What we haven’t seen yet is a breakthrough in how production IT works. Most of the DevOps activities that we see stop short of continuous deployment into production.
We see a few gaps that are the key inhibitors that must be addressed in 2016 for full production deployment.
1. DevOps tools need to be more enterprise grade and production ready
While DevOps tools might be great for development and test use, they are often not capable of handling large-scale uses that would be required for automated deployment into large production data centers. The DevOps software vendors that are working on scalable deployments and are developing interfaces beyond just programming interfaces “infrastructure as code” will capture the enterprise customers in 2016.
As an enterprise that is evaluating DevOps tools, I would suggest making sure that you test them in a large-scale production use case – even if you are just doing a pilot. The most developer-friendly tools tend to get adopted early by developers, but they may not be the tools that you want to build your operational deployments on.
2. DevOps needs to start looking beyond process and also look at how to replicate the production environment for development and testing
Replication would help ensure there is little or no change when deploying into production. The DevOps tool chain needs to have tools in the mix that can do “sandboxing” – create a replica of the production environment – including network security and hybrid cloud infrastructure if that is used in production. If all of the development, testing, and QA is done in a sandbox that is a high-fidelity replica of the production environment, then the risk of deployment will go down significantly.
Only a small set of software vendors offers this capability, but in 2016 it needs to become a regular part of every DevOps tool chain.
3. Automation needs to move beyond “infrastructure as code”
In 2015, the focus of most DevOps tools was automating the steps in the DevOps process. In order to fully implement the DevOps process, the tools need to enable automation of all parts of the process, not just the movement between the steps of development, test, QA and production. Automation needs to be an enabler of the DevOps process as a whole and this includes automating processes, applications and infrastructure from development all the way through to production.
Automated application deployment means much more than just pushing the software onto a server in production. Applications may get deployed onto clusters of virtual machines rather than individual servers. They may get deployed to different parts of a hybrid cloud depending on capacity. This needs to be part of the application development and testing process, rather than introduced as a new process in production.
In 2016, enterprises should be looking for end-to-end solutions for DevOps to ensure that they can implement the whole process. This might mean limiting the range of tools that they use to those that work well together and can automate the full DevOps process.
4. We still need the “easy button” for operations
All the DevOps work in development will not result in the kind of rapid deployment that we hear about at Netflix and Amazon if the process stops short of actual deployment. The operations team needs tools that they can use to manage production from the point of deployment through the life cycle of applications. “Infrastructure as code” DevOps tools are not very useful for this work because management in production cannot be done by writing scripts.
Enterprises need good automation that is easy to use for people who are less programming oriented. Production environments need to be delivered with simple self-service interfaces and easy-to-use automation for life cycle management.
In 2016, enterprises should work to meet the requirements of the “Ops” side of DevOps. Enterprises may also want to look for DevOps Automation tools that integrate seamlessly with the production operating environments like vCenter and AWS.
Joan Wrabetz is CTO for QualiSystems. Earlier she was VP/CTO for EMC’s emerging product division. Joan has over 20 years’ technology executive experience. She was founder/CEO of Aumni Data, CEO of Tricord Systems (now Adaptec), VP/GM at StorageTek, founder/CEO of Aggregate Computing (now Platinum Technologies) and held management positions at Control Data Corporation and SRI International. She was a BlueStream Ventures partner, on the board of many startups and holds multiple tech patents.