What is DevOps?
It is the practice used in deploying software systems. Here they bring together software development, testing, and operational aspects. The purpose is to build and launch high-quality software as quickly as possible. This makes the processes more efficient.
What is MLOps?
Machine learning operations apply the same principles of DevOps to the field of machine learning. Here the software product is substituted by the Machine learning model. The ML model goes through the same steps of development, integration, testing, deployment, and monitoring as DevOps.
Similarities in MLOps and DevOps
-The lifecycle steps of MLOps and DevOps are the same.
-In both, there is a continuous integration of codebase amongst the developers.
-Outcomes of both are top quality, faster updates and releases, and higher end-user satisfaction.
-Continuous delivery of the system into production
Differences between MLOps and DevOps
In development process
In DevOps, first, a software application is created, then it is deployed and runs through a series of tests. The process is repeated till the final product meets the required goal.
In MLOps, building the model involves the testing part, where the test data is cross-checked against the predictions. The cycle continues till the model has reached the expected accuracy.
The ML systems are highly experimental in nature with no guaranteed success.
Version control
It involves tracking and managing changes to the software product.
In DevOps, focus is not tracking the changes in the code and artifacts, making it less complicated.
In MLOps, there is a need to track training data, statistics, distribution, experiment run, and other elements. So it requires tracking the variables of each and every experimental run.
Monitoring
In DevOps, it involves monitoring the entire lifecycle from planning to deployment, which includes development, integration, testing, deployment, and operations.
In MLOps, the focus is on monitoring the machine learning model and its algorithm itself to identify accuracy defects.
Continuous training
This concept does not exist in DevOps. This is about automatically identifying events that require the model to be retrained into production due to performance degradation in the current ML system.
- References:
- https://www.analyticsvidhya.com/blog/2022/01/mlops-vs-devops-lets-understand-the-differences/
- https://towardsdatascience.com/mlops-vs-devops-5c9a4d5a60ba
- https://www.cioinsight.com/big-data/mlops-vs-devops/
- https://valohai.com/blog/difference-between-devops-and-mlops/
- https://www.launchableinc.com/devops-vs-mlops-differences-similarities
Please Don't Forget To Join Our ML Subreddit
Prathvik is ML/AI Research content intern at MarktechPost, he is a 3rd year undergraduate at IIT Kharagpur. He has a keen interest in Machine learning and data science.He is enthusiastic in learning about the applications of Machine learning in different fields of study.
