----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/13603/ -----------------------------------------------------------
(Updated Jan. 14, 2014, 7:08 a.m.) Review request for mesos, Benjamin Hindman and Jie Yu. Changes ------- Rebased. I'm considering simplifying the TimeSeries in this change since it's potentially solving a problem that we don't have. If others think it is too complex then I'll update it to be a simple circular buffer style TimeSeries (i.e. remove sparsification). However, it does have one nice property in that if the window capacity is not exceeded, then the current TimeSeries performs no sparsification. Repository: mesos-git Description ------- The first part of this change is the removal of the archive operation. This also exposes a TimeSeries data structure used by Statistics. The goal here is to provide reasonable historical information for high frequency time series that rapidly exceed capacity. TimeSeries allows a window of values to be stored, while "sparsifying" older values when the capacity is exceeded. See the documentation on TimeSeries for details. An alternative, more deterministic approach, is to use sampling on older values. Both approaches have their advantages and disadvantages. Diffs (updated) ----- 3rdparty/libprocess/include/process/statistics.hpp ce122a5eaa1aa9c09207cc8c9428136b681561cf 3rdparty/libprocess/src/statistics.cpp d4ba9f146f7b9b46525a0e27fbfb3d61a21a94fc 3rdparty/libprocess/src/tests/statistics_tests.cpp e6c9a1b776d6f3339f96898c3501d2a00c416006 Diff: https://reviews.apache.org/r/13603/diff/ Testing ------- Added tests to verify the truncation and sparsification of time series. Thanks, Ben Mahler
