I perform workforce analyse duties at work and have consistent performance issuers with my device. I spend ~5 hours a week waiting.
These devices are essentially thin clients. Mid grade i3's and 4GB of RAM.
I have analyzed the issues the best I can and have come up with the information that strongly suggests the performance issues stem from a lack of memory, not lack of processing power. My findings:
- Memory is always at 95%+ usage, it sporadically drops during idle times (I assume paging to disk?)
- CPU usage averages 5-10% throughout the day
- When performing a memory intensive task, page-faults skyrocket and so does CPU usage and drive I/O
- When opening a new browser tab, or any program that has been idle for a few minutes (even sticky notes) the page faults/s for that process spike (5,000 - 15,000 faults/s) and the application/tab hangs for a while before opening. Drive I/O also spikes during these times.
- Applications, once open, perform fine. The performance issues hit when switching to another application or tab.
I also made a VM at home with a single core from my i7 and 4GB of RAM. I ran into the exact same performance issues, I doubled the RAM while keeping the allocated cores the same and the hanging completely went away while performing the same tasks.
I've approached the IT director with these findings, and he is convinced that RAM will not increase performance, that the issue is solely CPU and drive I/O bound. My program manager is starting to think I'm a fool for trying to push for more RAM while the site IT guy says that's not the issue.
What can I do to definitively prove this is a memory bound performance issue?