lessfs v0.2.1 significantly faster

  • Andreas Lund
    Andreas Lund

    Finally had time this weekend to download and try 0.2.1, still using my old P3 1 GHz and 4K block size. Previously the performance would drop sharply when passing about 10 gigabytes, not anymore so you must have done something right :-)

    I am curious about the new tuning options CACHESIZE and COMMIT_INTERVAL. For large sequential writes, this causes rsync to work for 30 seconds, pause for about 30 seconds, work another 30 seconds etc. instead of just churning along as it used to. This results in an overall average speed of 250-300K/sec.

    As large sequential writes effectively turns caching into overhead, would you recommend I reduce cache size, decrease the commit interval, or both? Or should I perhaps increase the commit interval and just let the cache overflow and commit as needed?