> On April 18, 2014, 12:42 p.m., Ben Mahler wrote: > > 3rdparty/libprocess/include/process/statistics.hpp, line 74 > > <https://reviews.apache.org/r/20047/diff/6/?file=561284#file561284line74> > > > > Looks like this could leverage the 'Percentage' abstraction I mentioned > > in some earlier patches, because there's some ambiguity in the caller as to > > whether they need to call this with [0..1] or [0..100]. Could we add a TODO > > for that? > > > > But taking a step back, do we need to keep around all of this data? > > Could we just pre-compute a 'Percentiles' struct in the constructor which > > provides p50, p90, p99, p99.9, p99.99? We'll likely want to expose a > > consistent set of percentiles for all metrics anyway, right? > > > > Would be great to add a TODO that captures our intent to eventually add > > real-time histograms as well if it proves to be needed for efficiency. > > Dominic Hamon wrote: > As this is a general purpose abstraction, allowing flexibility from the > API is key. We do eventually expose a consistent set of percentiles as the > MetricsProcess only pulls out a subset (just p50, p90, p99). > > I'd prefer to keep this class as general as possible as it may find uses > elsewhere. > > Ben Mahler wrote: > Do we need it to be general purpose though? This seems nice, but it's > probably a bit too general for our needs? > > A fixed-percentiles 'Percentiles' struct is nice because it will force us > to consistently expose them, right?
added in latest patch. I considered making Statistics the Percentiles struct (as it were) but this makes every percentile an Option<T> with associated get() on access. This patch has the Percentiles struct returned as an Option and then every value within it a concrete T. This makes the API cleaner. - Dominic ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/20047/#review40803 ----------------------------------------------------------- On April 18, 2014, 12:54 p.m., Dominic Hamon wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/20047/ > ----------------------------------------------------------- > > (Updated April 18, 2014, 12:54 p.m.) > > > Review request for mesos, Benjamin Hindman and Ben Mahler. > > > Bugs: MESOS-1036 > https://issues.apache.org/jira/browse/MESOS-1036 > > > Repository: mesos-git > > > Description > ------- > > see summary > > > Diffs > ----- > > 3rdparty/libprocess/Makefile.am d707ad759dacd16e0177e14f1bf5ece9e4ce2491 > 3rdparty/libprocess/include/process/statistics.hpp > a4f1db3a8a219c39193a1d237477f0350e47e681 > 3rdparty/libprocess/src/process.cpp > 9654c0437edb43cff65dbefdf08dee9e18ef96ab > 3rdparty/libprocess/src/statistics.cpp > 75aac4074d33cb5054da6c8b0bd4a890c2eaf80e > 3rdparty/libprocess/src/tests/statistics_tests.cpp > 3521bd565dae8fcbba464f2539b3b14a37a037f0 > 3rdparty/libprocess/src/timeseries.cpp PRE-CREATION > > Diff: https://reviews.apache.org/r/20047/diff/ > > > Testing > ------- > > make check > > > Thanks, > > Dominic Hamon > >
