The Search for a Silver Bullet. Again…
Every few years there is a new technology that takes the IT world by storm. Virtualization, cloud, IaaS/PaaS/SaaS, containers, microservices, big data, machine learning, deployment automation. The list apparently can go on as evidence of continuous creativity and entrepreneurship. And every time a new concept, name, technology shows promise and its adoption starts the IT community announces that this time the new technology will make traditional operations issues go away. Incident recovery becomes a piece of cake as you just re-deploy an older working virtual machine image. Configuration issues go away as you just rely on a catalog of standard configurations provided through infrastructure-as- a- service. Root cause analysis becomes easy as you automatically mine big operational data detecting anomalies through machine learning algorithms. Vendors developing these technologies strongly contribute to this impression by the force of their conviction and power of their education of the market.
The Never-ending Hype Cycle
I have been part of this cycle of hope and despair for the last 25 years. Gartner even formalized this transition of expectations in their so-called “hype cycle”, because hype always fades away. While you may discover that the new technology is very powerful, it still does not solve all the world’s problems and actually introduces a few new ones. For example:
- Incident Recovery: While you can easily roll back your environment to a previous state to recover from an incident in a virtualized/cloud world, however, you need to know what point to recover to, which is not obvious. An incident could be the result of a change made a few versions of virtual images ago. Furthermore, what if your data schema changed and you already acquired new data in the new format? What if another application relies on the new APIs you introduced and recovery downgrades the API?
- Configuration Issues: Standardizing infrastructure configurations drastically reduces the amount of configuration issues, improving stability and reliability of the infrastructure and dependent systems. However, for an enterprise, you most probably will need to support a share of custom configurations, unless you refactor all of your hundreds and thousands of applications.
- Root Cause Analysis: You can automatically detect anomalies in IT operational data and investigate them to understand and predict performance and availability issues, however anomalies are not necessarily the root cause. Frequently they are just a symptom that an issue is present.
Here We Go Again
When I meet IT architects in banks, insurance companies, healthcare organizations, large retailers and other enterprises, they routinely announce that they are in the process of transition to cloud, containers, microservices etc. They declare that the new architecture will address all the needs for stability and performance. The need for visibility and control will disappear as the new architecture will self-manage, self-heal and self-report, protecting itself from the IT operations staff that can mess it up.
Contrastingly, operational staff frequently cringe upon hearing this, having already experienced a few iterations that were supposed to bring them closer to the holy grail of “no-ops”. Nevertheless, their job becomes even more challenging than ever.
Waiting for the Silver Bullet
I am all for new technology. Constant evolution makes IT more powerful, and more efficient. Now we can deliver in minutes what would take months to develop, just 15 years ago.
However, these new technologies will be more effective if we don’t blindly announce that they can solve all IT’s known issues. Automation needs checks and balances, flexibility needs visibility, IT operations will involve the human factor until AI finally takes over.
To cut a long story short, IT would be much more effective if they would design and implement operations controls, checks and analysis as an integral part of any new technology rollout. One day a silver bullet may come, but for the meantime let’s stay prepared.