Re: PMDA with 2 levels of inst

["Frank Ch. Eigler" <fche@xxxxxxxxxx>]

--text follows this line--

(sorry, necrothread followup, rescued a lost draft)

kenj wrote:

> [...]  The underlying PCP data model supports only linear instance
> domains, [...]

Have we encountered modeling situtations where the downsides of the
workarounds you listed in detail are hard to bear?  I can think of a
few cases, where we're sacrificing expressive power, like stats that
naturally cross separate domains like per-process per-disk or
per-network-peer traffic.  At what point should we investigate
generalizing to multiple indoms per metric?

- FChE