> On Oct. 10, 2017, 2:39 p.m., Stephan Erb wrote: > > src/main/java/org/apache/aurora/scheduler/storage/backup/Recovery.java > > Line 143 (original), 146 (patched) > > <https://reviews.apache.org/r/62873/diff/1/?file=1851364#file1851364line148> > > > > Bill, totally unrelated question: Has anyone of you ever experimented > > with the TCompactProtocol rather than the TBinaryProtocol? > > > > I have heard rumors that it should be smaller while providing similar > > performance. So, it might be worth a try for our IO bound operations.
As it turns out, i recently experimented with this on my laptop to see how it would affect backup sizes. IIRC it only shaved a few percent off the total backup size, and the difference disappeared when the backup was gzipped. Minimal enough that i did not investigate further. - Bill ----------------------------------------------------------- This is an automatically generated e-mail. To reply, visit: https://reviews.apache.org/r/62873/#review187594 ----------------------------------------------------------- On Oct. 10, 2017, 1:35 p.m., Bill Farner wrote: > > ----------------------------------------------------------- > This is an automatically generated e-mail. To reply, visit: > https://reviews.apache.org/r/62873/ > ----------------------------------------------------------- > > (Updated Oct. 10, 2017, 1:35 p.m.) > > > Review request for Aurora and Jordan Ly. > > > Repository: aurora > > > Description > ------- > > This reduces the memory burden of loading a backup for recovery. Previously, > the backup file would be fully loaded into a `byte[]`, which may be very > large and fail to allocate. > > > Diffs > ----- > > src/main/java/org/apache/aurora/scheduler/storage/backup/Recovery.java > b74de9b51fd01788a3970e56a534a3b9adcd8863 > > > Diff: https://reviews.apache.org/r/62873/diff/1/ > > > Testing > ------- > > > Thanks, > > Bill Farner > >