The Storage benchmark score is calculated as follows:
Storage test score = 44 × geomean(R1 ,R2 × 2, R3, R4, R5 × 2, R6, R7, R8, R9 × 2, R10 × 2, R11 × 2, R12 × 2) Where: R1 = The arithmetic average of the read times from task 1 R2 = The arithmetic average of the read times from task 2 R3 = The arithmetic average of the write times from task 3 R4 = The arithmetic average of the write times from task 4 R5 = The arithmetic average of the read times from task 5 R6 = The arithmetic average of the read times from task 6 R7 = The arithmetic average of the write times from task 7 R8 = The arithmetic average of the write times from task 8 R9 = The IOPS measured in task 9 R10 = The IOPS measured in task 10 R11 = The IOPS measured in task 11 R12 = The IOPS measured in task 12
The weights used in the score formula are based on studies(1,2) that suggest that the most common I/Os issued by applications are:
- Random reads on the internal storage for loading application binaries, (represented by workload task 2).
- Sequential reads on the external storage for multimedia files playback, (represented by workload task 5).
- Database access for application internal data and Android contents, such as the dial log, contacts, web browsing logs, and multimedia databases, (represented by workload tasks 9, 10, 11, and 12).
Secondary results
The table below shows how each secondary result is related to a workload task, its unit, and its explanation.
| Task | Result | Unit | Explanation |
|---|---|---|---|
1 | Internal sequential read | MB/s | The median bandwidth over the results measured from the repeats of task 1 |
2 | Internal random read | MB/s | The median bandwidth over the results measured from the repeats of the task 2 |
3 | Internal sequential write | MB/s | The median bandwidth over the results measured from the repeats of the task 3 |
4 | Internal random write | MB/s | The median bandwidth over the results measured from the repeats of the task 4 |
5 | External sequential read | MB/s | The median bandwidth over the results measured from the repeats of the task 5 |
6 | External random read | MB/s | The median bandwidth over the results measured from the repeats of the task 6 |
7 | External sequential write | MB/s | The median bandwidth over the results measured from the repeats of the task 7 |
8 | External random write | MB/s | The median bandwidth over the results measured from the repeats of the task 8 |
9 | SQLite read | IOPS | The number of records read per second during task 9 |
10 | SQLite update | IOPS | The number of record updates performed per second during task 10 |
11 | SQLite insert | IOPS | The number of record insertions performed per second during task 11 |
12 | SQLite delete | IOPS | The number of record deletions performed per second during task 12 |
- Lim. S. H., Lee S. & Ahn W. H. (2013) Applications IO profiling and analysis for smart devices. Journal of Systems Architecture, 59, 9, 740-747.
- Lee K. & Won Y. (2012) Smart layers and dumb result: IO characterization of an Android-based smartphone. EMSOFT '12 Proceedings of the tenth ACM international conference on Embedded software. ACM New York.