raghavm23 / comparison-of-nosql-db-using-ycsb

In this project, the architecture and performance of two NoSQL databases ie. MongoDB and Cassandra are compared using Yahoo Cloud Service Benchmark

MongoDB vs Cassandra (performance comparison)

Modern day applications create exceptionally large volumes of ever increasing data and ever changing data types. Hence it is not feasible to store such data in the conventional relational databases as they are not optimised for horizontal scaling. This paves the way for NoSQL databases and in this project we compare the architecture and performance of two such NoSQL databases ie. MongoDB and Cassandra.

Yahoo! Cloud Service Benchmark Framework

Developed by Brian Frank Cooper

The YCSB tool consists of a ycsb-client which generates the workload and the defined workloads which are the scenarios to be executed by the client.

Performance Test Plan

Two NoSQL DB compared- MongoDB and Cassandra

Two predefined workloads A & B are chosen for the performance testing Workload A : Update heavy workload – 50/50 reads and writes Workload B : Read mostly workload – 95/5 reads and writes

Five different number of operations chosen for testing are 100000, 250000, 500000, 750000 and 1000000**

Three times each test is performed for accurate and consistent results

--60 total iterations--

Conclusion

MongoDB provides superior performance than Cassandra in my testing scenarios. The overall throughput is much higher for MongoDB, sometimes attaining more than 5x of Cassandra. MongoDB also exhibits much lower latency in both the workloads across all operation counts.