Achieving Highperformance Functiondriven Distributed Computation Across Containerized Infrastructure Ecosystem

One of the most well-known and powerful parallel computing models for big data is MapReduce. Even now, it is still being used in the big data systems ecosystem, and it is now changing in the cloud-native environment. One of the most promising types of cloud-native technology is serverless computing because it is flexible and easy to use. Adding MapReduce big data processing support to the Serverless environment could make Serverless even better. However, because the underlying system architecture was different in the two environments, there are three problems that will come up when running MapReduce jobs in the Serverless environment. First, the scheduling strategy can't make the most of all the resources that are available. Second, accessing, reading, and changing Shuffle index data that is stored in the cloud is slow and expensive. Lastly, the Input/Output (I/O) requests that involve storage on cloud architecture show a long-tail latency problem. To fix these problems, we suggest three ideas for the MapReduce parallel processing framework that would work in a Serverless environment. We have demonstrated experimental results indicating that, in comparison to state-of-the-art competing work, our approach achieves a 25.6% reduction in job execution time and a 17.3% decrease in the average cost of job execution.