[Remember that FIO is a tool used for benchmarking and stress testing IO. It’s a generic tool that can be used to test different types of drives, IO operations and IO engines. FIO is an open source tool that can be downloaded from here. You can install it after compiling the source.]
Our benchmarking methodology for FIO is described here. The new generation of compute optimized instances are ‘c3’ instances. They have SSD drives for their instance store.
Details of all instances in the ‘c3’ family can be found here. Similar to our previous disk bandwidth benchmarking, we benchmarked the instance store. We made sure the load is evenly spread on all the instance stores.
The following table shows the results for FIO disk bandwidth performance.
Here is a chart indicating the disk bandwidth.
This chart gives us an idea of the disk bandwidth per dollar per hour. We also see that c3.large offers great performance per dollar amount.
We also did a break up of FIO performance based on the type of I/O operations. The table and the chart below describes the disk performance for different types of I/O.
In the end, we see that the c3.large benchmarked really well for sequential write when compared to other instances, except for c3.8xlarge. It would be even more interesting to further investigate why we got better performance on c3.large when compared to other instance types in the ‘c3’ family.