tag:blogger.com,1999:blog-3617908412741200959.post6953569363481431773..comments2024-03-18T22:03:43.359-07:00Comments on sql.sasquatch: SQL.Sasquatch, SQL_Sasquatch, SQL-SasquatchSQL_Sasquatchhttp://www.blogger.com/profile/13470482959972282429noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-3617908412741200959.post-20354428965203556922013-09-13T20:16:50.388-07:002013-09-13T20:16:50.388-07:00Hello obulay! Replication of a database can take ...Hello obulay! Replication of a database can take place at several different levels - database, filesystem/LVM (IBM LVM mirroring is the one I've worked with most at the LVM layer), storage. On each of those three levels there is the possibility of synchronous or asynchronous replication. I like to start negotiations with asynchronous database replication, and ask for solid reasons to vary :)<br /><br />Synchronous replication has the most significant potential for performance impact, because even if the local and remote systems can both absorb replication activity resource utilization without breaking a sweat, the round trip wire time between systems and acknowledgment necessary for synchronous replication can impose significant additional latency to every write operation on the local system.<br /><br />Although asynchronous replication can avoid some of the performance impact associated with synchronous replication, there's still a lot of activity involved. And asynchronous replication ALWAYS implies some maximum tolerance for latency between the systems - most typically this is represented by the finite capacity of a log location on the local system. Once that tolerance is reached, either the replication relationship needs to be severed, or activity on the local system may be suspended or transition into something similar to the one-in, one-out scenario that comes along with cache saturation.<br /><br />Anyway... beyond the basics, it requires details of at least the local system, remote system, connecting network, and replication mechanism to determine where replication activity and resource utilization could present itself as overhead leading to performance impact. SQL_Sasquatchhttps://www.blogger.com/profile/13470482959972282429noreply@blogger.comtag:blogger.com,1999:blog-3617908412741200959.post-63022766704413126302013-09-11T17:41:41.786-07:002013-09-11T17:41:41.786-07:00Hi
I am new to replication and wonder what is rep...Hi<br /><br />I am new to replication and wonder what is replication overhead. Where are the performance issues are introduced? <br />by the log reader reads?<br />by distribution database writes?<br />by subscribers/Anonymousnoreply@blogger.com