> > On 4/16/2014 1:44 PM, Zuckerman, Boris wrote: > > Typically File System services do not offer semantics of transactional > > isolation. Attempts to add that took place and were rejected. > > Therefore, we are speaking only about some potential performance > > penalty, right? > > Not at all. This has nothing to do with transactions; it is simply a question of being > able to identify what process is causing all of the disk IO, like via iotop. By counting > writes in the process that initiates the writeout of dirty pages, rather than the process > that dirties the page, it renders the write io tracking largely inaccurate to the point of > being useless at times, since often the process dirtying the pages, even with an actual > write() system call, is not the process that initiates the writeout. > [bz:] In such case I'd rather have lighter implementation of caching layer, than ability to track disk IOs per process. IOs per process can be tracked by other tools... -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
