Mobile

Real-Time

Enterprise

Embedded

Mimer SQL Real-Time - Benchmark

Introduction 

Performance Benchmark

In this benchmark the worst-case response-times of read and write operations using the database pointers provided by the Mimer SQL Real-Time API. The benchmark has been performed on two hardwares, (i) an Intel Core2Duo 2.33GHz and (ii) and an embedded processor, Intel Xscale 533 MHz.

Predictability Benchmark 

In this benchmark, the predictability and efficienct of Mimer SQL Real-Time compared to a non real-time database is measured. The benchmark executes typical real-time control code where several control modules communicate with each other in a control feedback loop. Three different configurations has been used to show the impact of the predictability. The configurations are:

• Communication is performed without a database, instead it uses shared variables protected by Pthread Mutexes. This is the way many control applications are implemented today.
• Communication is performed using a standard SQL client/server database.
• Communication is performed using Mimer SQL Real-Time.

The graph below clearly show that the Mimer SQL Real-Time does introduces almost no extra overhead as well as maintain the real-time predictability of the system. In-fact, Mimer SQL Real-Time only introduces and overhead of 3% in this benchmark.

On the other hand, using a non-real-time client/server SQL database has a clear negative impact of the performance and predictability of the application.