MySQL Forums
Forum List  »  NDB clusters

Re: NDB cluster read path and write path
Posted by: Mikael Ronström
Date: August 29, 2017 03:45PM

The way it works is the following, the in-memory data is updated
in-memory and REDO logs are sent to disk within a second. REDO logs
together with checkpointing ensures that all this data is durable
in several memories at commit time and also on durable media within
about 1 second after the commit.

The in-memory data stays in memory all the time while the node is up.

There is also an option to have disk-based columns. These columns cannot
have indexes. These columns are also updated in-memory at commit time
and REDO logs are sent to disk within a second after commit. The difference
is that on-disk columns can be paged out from memory when the pages are not
used.

This is the way NDB works, so no special configuration is needed. The
in-memory or on-disk column is decided when you create the table, if
the column is declared with STORAGE DISK it will be on-disk.



Edited 1 time(s). Last edit at 08/29/2017 03:45PM by Mikael Ronström.

Options: ReplyQuote


Subject
Views
Written By
Posted
Re: NDB cluster read path and write path
521
August 29, 2017 03:45PM


Sorry, you can't reply to this topic. It has been closed.

Content reproduced on this site is the property of the respective copyright holders. It is not reviewed in advance by Oracle and does not necessarily represent the opinion of Oracle or any other party.