Correlate graph plottings to usage snapshots (and X-axis time intervals)

Started by Coldblackice, October 29, 2020, 10:02:55 PM

Previous topic - Next topic

Coldblackice

It would be awesome if there was an ability to select/click/highlight points or areas on the graph and then have a floating popup reveal information about what was occurring at that point in time.

For example, if I look at the graph and see a "V" shape cut into the Responsiveness plot line, it would be highly useful if I could mouse over to that "V" on the graph, click, and then a floating popup appears showing what the top-5 CPU-using processes were at that time (or RAM/commit, I/O, Network usage, etc).

Also useful might be graphing timestamps on the X-axis to help correlate what was occurring at those moments. This usefulness would be compounded if there was an ability to choose the time-increments, like 30-sec intervals, 5-min, 15-min, 30/60/etc. Obviously the graphing would change along with these changed intervals.

Jeremy Collake

I agree that would be useful, and have had similar functionality on the radar. I'll certainly keep it in mind. Thanks for the feedback!
Software Engineer. Bitsum LLC.


Coldblackice

Circling back on this, correct me if I'm wrong, but it seems this has been implemented, resulting in the occasional yellow vertical graph columns marking a particularly resource-aggressive process at the time -- awesome!

By chance are there any settings to tweak the parameters for when this should occur? If not, would it be possible to implement?

(*incoming wordy brainstorm*)

It'd be useful having the ability option to see more data visualized on the graph (which I find myself relying on more often, glancing at it and the Actions log when I notice a possible hiccup), such as more of these program marker bars, given that the majority of the time the graph is largely empty/flatlined and thus unused, at least until something happens. This space could instead be utilized more if there was an option to define markers that get imprinted on the graph, like what constitutes a resource-aggressive process to be printed on the graph.

For example, one setting/rule could be "If a process uses X CPU for > Y seconds, imprint on graph", or "If a process gets Pro-balanced X times...", "...if peak working set tops X", "...if X PB's overall happen within a minute", etc.

Probably getting ahead of myself here (just brainstorming), but it could even go further with differently colored columns indicating a particular resource/metric being overused at the time, e.g. a green column for CPU over-use, blue for RAM, orange for GPU, etc. Potentially overlapping bars could be gradiented left->right to remain distinguishable, like if two processes both used heavy CPU at the same time. Or bars could default to lower opacity, making them slightly transparent and with outlines, resulting in slightly darker color overlap.

Another possibility which would be simpler from a dev perspective is time indicators along the graph's X-axis. Since the graph moves at a static pace, X-axis time indicators could help visually correlate with actions in the log, like who was responsible for the memory spike that happened X minutes ago. And this would also help correlate troubleshooting beyond PM, like with Windows Event Viewer, honing in on a 10-minute block of time where the graph showed resource usages were spiking. Process column markers on the graph would be ideal, but the time indicator would also help in addition to this, correlating the time surrounding with the Actions log.

Sure, one could technically do this looking at the hard log data alone, but given the graph has a lot of underutilized space or loosely static/uneventful data most of the time (obviously dependent on one's hardware), I feel having the option to enable more data displayed on the graph -- particularly aggressive processes -- would be highly useful.

But nonetheless, I'm just brainstorming off the top of my head, not criticizing or anything. Nobody's complaining by not having this, I just thought it would be a useful feature addition to Process Lasso IMO.

Jeremy Collake

The highlights you see are ProBalance events, and have been there all along. They were improved a bit in v10 though.

I like the idea of showing more useful data about resource use spikes, or other process events. I also like the idea of making this configurable by the user. Time period indicators, for sure. There are some other neat ideas related to this that I won't delve into here.

This sort of graph improvement has long been on the radar, but it obviously takes some real design and labor to pull off, so hasn't made its way to the top of the planned feature list. I'll absolutely keep it in consideration. It may very well happen, but I don't want to commit to it right now.

Thanks for the feedback!
Software Engineer. Bitsum LLC.

Coldblackice

Awesome, sounds good! And no problem. I don't doubt it'd be a pickle to implement, I don't take for granted the work it must take to juggle all the varied development needs/priorities on your plate.

Anyway, looking forward to a hopeful/eventual release!