Impact of TLS Overhead on Segmented Network for Cloud Database Systems

Abstract

Cloud database serves flexible, affordable, and scalable database systems. Even the cloud database is secure with transport layer security (TLS), but the performance overhead that TLS introduces when executing operations on one of the major No SQL databases: Mongo DB in terms of latency. To explore TLS performance overhead for Mongo DB, we performed two tests simulating common database usage patterns. We first investigated connection pooling, where an application uses a single connection for many database operations. Then, we considered one request per connection in which an application opens a connection, executes a process, and immediately closes the connection after completing the operation. Our experimental result shows that applications that cannot endure significant performance overhead should be deployed within a properly segmented network rather than enabling TLS. Applications using TLS should use a connection pool rather than a connection-per-request.