Considering this as feature request, changing the status to Untriaged

Thanks.  I think we'll need to be more precise about what to measure and what it should be compared against, as I mentioned on the review of https://codereview.chromium.org/2694363007/.

You were correct in noting that there's no way to get memory or CPU metrics for just part of a process, so we're limited to considering the impact that OOPIFs have to the processes they're in.  Today, we're doing A/B comparisons for modes that have OOPIFs enabled and disabled, such as the SiteIsolationExtensions field trial.  That's a global view, though, and doesn't answer what the size of processes with OOPIFs are.

There's several cases involved:
1) Typical renderer processes with no OOPIFs.
2) Processes that only contain one (or more) OOPIFs.
3) Processes that were created for an OOPIF but also have main frames (e.g., if an OOPIF opens a popup).
4) Processes that were created for normal pages but had an OOPIF placed inside them (e.g., an extension process in --isolate-extensions mode).

You could imagine comparing cases 1 vs 2+3.  (We would want to be careful to compare 1 vs 2+3 and not 1+2+3 vs 2+3.  In other words, collect a value for processes with no OOPIFs and a comparable value for processes with at least 1 OOPIF.)

However, I feel like including 4 in the second group would distort the meaning of the query, and I expect 4 is pretty common now that --isolate-extensions is enabled.

I suppose an alternative would be comparing processes that *only* contain OOPIFs with all other processes.  That might help answer questions about whether such processes tend to be smaller or less resource intensive.

Nick, what are your thoughts here?