On Wed, Jun 08, 2016 at 05:13:09PM +0000, Rodriguez Betancourt, Esteban wrote: > I was performing some profiling on OVSDB, when inserting a lot of > rows. The callgrind results shows that the SHA1 calculation takes near > 10% of the time within our test (the whole file writing, including > SHA1, takes like 20%). > > We want to know further about why that checksum is need, in order to > decide how it could be optimized (we are considering: using processor > specific instructions, changing the algorithm or removing it at all).
The checksum is there to make writing the log atomic. Without the checksum, it would be hard to tell whether data read from the file after a header line was actually the data that was meant to be read or just junk that happened to be on disk when there was a power failure or data corruption. A weaker checksum could serve the same purpose. > By the way, I would want to know if there are any updates on the > review of the IDL Compound Indexes (pull request #127). I guess I lost track. I tend to forget about pull requests because they don't show up in patchwork. Sorry about that, I'll try to get to it soon. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
