How to measure performance improvement over unbuffered file I/O?
posted 10 years ago
I've a piece of code that does quite a lot of file I/O. The application is running on Solaris. Gotten a lot of high system CPU utilization and read somewhere that this is due to unbuffered I/Os. So I rewrote the code and uses buffers for it.
How do I quantitatively shows that the buffered codes is better than the unbuffered codes? I could simply roll the new codes out to production, but thats really not the ideal case since I don't have any hard figures to show that the codes would fix the high system CPU usages.