Comments on: Using BASE instead of ACID for scalability http://www.xaprb.com/blog/2008/07/23/using-base-instead-of-acid-for-scalability/ Stay curious! Tue, 06 Jan 2009 08:46:08 +0000 http://wordpress.org/?v=2.6.2 By: Kunal Jain http://www.xaprb.com/blog/2008/07/23/using-base-instead-of-acid-for-scalability/#comment-14937 Kunal Jain Tue, 05 Aug 2008 12:02:53 +0000 http://www.xaprb.com/blog/2008/07/23/using-base-instead-of-acid-for-scalability/#comment-14937 Talking about Replication I read your book High Performance Mysql. I have One Question. What if one on a Master server have different Databases (Example DB1 DB2 DB3 ..) so how do we replicate all those Databases on the slave server. I have two Database servers one act as a Master and the Other as Slave. What i have is Multiple database running on the Master and i am able to replicate only one. how do i replicate all the databases on the slave server. Talking about Replication I read your book High Performance Mysql. I have One Question.
What if one on a Master server have different Databases (Example DB1 DB2 DB3 ..) so how do we replicate all those Databases on the slave server.
I have two Database servers one act as a Master and the Other as Slave. What i have is Multiple database running on the Master and i am able to replicate only one. how do i replicate all the databases on the slave server.

]]> By: Ryan Lowe http://www.xaprb.com/blog/2008/07/23/using-base-instead-of-acid-for-scalability/#comment-14898 Ryan Lowe Thu, 24 Jul 2008 01:54:24 +0000 http://www.xaprb.com/blog/2008/07/23/using-base-instead-of-acid-for-scalability/#comment-14898 Indeed, as you said, this is a proven approach used by many "Web 2.0" sites (FaceBook, Ning, LinkedIn, etc...). Another dimension of this method of scaling is <a href="http://clarification.wordpress.com/2008/06/05/diagonal-scaling-and-the-law-of-diminishing-returns/" rel="nofollow">diagonal scaling</a>. Most sites that choose the BASE approach ignore the "E" (Eventually Consistent) aspect, in that the application does not completely "roll back" a multi-partition transaction if any part of it fails. This is often by design, especially for non-financial applications. As the DBA for a Web 2.0 site, I have found that a generally-acceptable solution for this is to have what we call a "Sanity Crawler"; essentially a script that continually looks for transaction completeness based on a pre-defined rules set (i.e. if a "Friendship Transaction" has not completed after N units of time, it will attempt to complete it or roll it back, based on the rules. If the transaction cannot be "fixed", a diagnostic message will be logged and developers will debug it). Indeed, as you said, this is a proven approach used by many “Web 2.0″ sites (FaceBook, Ning, LinkedIn, etc…). Another dimension of this method of scaling is diagonal scaling.

Most sites that choose the BASE approach ignore the “E” (Eventually Consistent) aspect, in that the application does not completely “roll back” a multi-partition transaction if any part of it fails. This is often by design, especially for non-financial applications. As the DBA for a Web 2.0 site, I have found that a generally-acceptable solution for this is to have what we call a “Sanity Crawler”; essentially a script that continually looks for transaction completeness based on a pre-defined rules set (i.e. if a “Friendship Transaction” has not completed after N units of time, it will attempt to complete it or roll it back, based on the rules. If the transaction cannot be “fixed”, a diagnostic message will be logged and developers will debug it).

]]>