Need for a product to handle Past and Future Batch workload types

As far as i am concerned Batch processing is the best example for need of elasticity of infrastructure. Whether it is ETL, File processing, Event based, Time based or Ingress request based. Lot of Organizations have full set of VM's hogging on to the computes to run a job once in a day, week or month. Lot of companies have mainframe batch jobs, windows task schedulers, linux Cron jobs, in house job schedulers etc.,
If you are looking to future, K8's Cron job, Volcano or Kube Batch can offer nice features to maintain batch job workloads in Kubernetes world. Serverless can be an alternative too. But the question is what is the difference between 0 VM's vs 1 worker VM. Because you are NOT going to save a lot by going to Zero. And if you are using something like Azure Functions as your Serverless strategy you need to understand, if you are really leveraging consumption model or NOT. Don't come to the assumption if you are using Azure Functions you are serverless.
The questions is how to transform thousands of batch jobs you have on VM's currently into a K8's or serverless without doing it from the scratch. I feel there is a niche here. If we have a product / Offering that can seamlessly migrate from VM based batch workloads into K8's / Serverless that will be great saving and step in the right direction for lot of companies in Erasing tech debt.
The product should give an option to control all the batch work loads in the org to be controlled through Same portal with an abstraction layer. That way you are still using the same portal if you are running a mainframe job or Windows task scheduler. Then the trick is to extend that portal functionality to K8's jobs as well. Underlying execution plane could be whatever you chose (K8 cron jobs or Kube batch).
Now users can run their jobs through the portal to their normal execution engine and beta test on K8's execution engine , get comfortable with it. Later they can shutdown their jobs in Linux VM and start running them in K8's. This way there will be seamless adoption and transition. This will help organization to entice more developers into K8's world.

Comments

Popular posts from this blog

Dhirubhai Notorious or Famous???????

Sri Sri kavitha

Excerpt from the book "Made to Stick"