Pravega: Performance impact analysis with Connection Pooling

In this paper we perform analysis on Pravega distributed streaming storage with and without connection pooling. When connection pooling is enabled, there will be a restriction of number of open connections for data transfer with Pravega cluster. Storage Benchmarking Kit is used to pump data of various sizes to the Pravega cluster running as docker containers with and without enabling connection pooling. The performance numbers are captured in CSV files which are then fed to sbk-charts which will produce simple charts which can be used for detailed analysis of each tests performed. All the tests are performed on two EC2 instances of type t2.xlarge on AWS Cloud infrastructure. Performance impacts are analysed with the generated charts for each test result which leads us to conclude on the performance gap when applying connection pooling during client applications integrations with Pravega server.