Analyze services for greediness

[Coldblackice]

Could we get a feature that keeps tabs on services, keeping stats on which services are using the most I/O, CPU, memory, etc. over time?

The problem with using something like Task Manager is that it only really shows you that one moment in time. If a process or service is getting greedy in short bursts over time, its resource usage adds up while not being easily detectable.

[edkiefer]

That is what Insights is for, showing history of processes that got restraint , which had to use X CPU over time in back-ground .
no it doesn't show CPU% values directly but maybe somethings could be expanded .

You can also setup watchdog to log when X processes goes over X CPU% to, or memory use .

[BenYeeHua]

For processes, it did recorded on the log and also Insights, but for the services..
I think for the services, it need many resource to support it, so far I know by using Process Explorer, you can check which threads are own by which service, and just stop it.

But if you keep monitoring it(as there are also a lot of svchost.exe), it will need a lot of CPU usage to do that.....

[Coldblackice]

That'll show the general "svchost.exe", however, multiple services can be stacked onto the same svchost.exe. But I'll check Ben's tip out -- I didn't know Process Explorer showed that.

[Jeremy Collake]

Lasso also shows you what services are hosted by a svchost.exe instance in brackets (group name in parenthesis). Gotta expand that column a bit.

An historical view of system activity is high on my project list right now. I can't say more right now.

[Coldblackice]

Lasso also shows you what services are hosted by a svchost.exe instance in brackets (group name in parenthesis). Gotta expand that column a bit.

An historical view of system activity is high on my project list right now. I can't say more right now.

Cool x2, thanks!

And have you considered the possibility of opening Process Lasso for more developer help? Perhaps even if just "externally", like letting a couple of us work on a module or library to do do/accomplish something -- helping to offload some of the workload from you, but while still leaving you with control to the main source, and ability to pick and choose what you want to use or include.

Perhaps similar to a plugin-type system. Obviously it wouldn't get down to the nitty-gritty low-level OS/API calls, but perhaps something more high level, like stats tracking or automated searching the PL-scraped data for for trends/patterns/etc.