On Fri, 2017-11-24 at 11:58 -0800, John R Pierce wrote: > On 11/24/2017 11:43 AM, Michael Nolan wrote: > > There are so many different levels of caching going on--within > > Postgresql, within the OS, within a disk array or SAN, and at the > > individual drive--that there may no longer be a meaningful way to > > perform this measurement. > generally, power cycling the server will flush all the hardware > caches AND the OS cache.
Given that a real-world application will almost never experience an empty-cache scenario I question the usefulness of clearing the cache(s). I would capture transactions from a production database in order to create a load test that mimics real-world load. -- Meetings Coordinator, Michigan Association of Railroad Passengers 537 Shirley St NE Grand Rapids, MI 49503-1754 Phone: 616.581.8010 E-mail: awill...@whitemice.org GPG#D95ED383 Web: http://www.marp.org
