Re: Question about snapshot speed

On Wed, 2007-09-05 at 16:47 +0200, Audrey Bourke (MI/EEM) wrote:
> Hi,
>
> I am wondering if anyone can give me times or relative speed performance
> of the MySQL Cluster snapshot facility.
>
> What I am trying to find out is if the snapshort works quickly enough to
> be considered to be of an exact time, OR, if the snapshot time is more
> of a "window", i.e. takes minutes - hours in which time changes could
> have occured.
>
> I understand these numbers will be affected by size of database etc. But
> if someone could give me timings & db size, I could estimate
> performance.

We write a REDO log. A global checkpoint is just flushing the REDO log to disk. i.e. making the transactions disk persistent. This is configurable and typically happens every second or two.

If we kept the REDO log since the beginning of the database, we could just replay it (i.e. every change to the database ever) to get back to the current state. Obviously, this would take a very long time. So we have an "optimisation" where we write a Local Checkpoint (a copy of the current data) which enables us to cut the tail of the REDO log. This means recovery can take a sane amount of time.

LCPs are all configurable and the quicker they happen, the less REDO log is needed.

-- 
Stewart Smith, Senior Software Engineer
MySQL AB, www.mysql.com
Office: +14082136540 Ext: 6616
VoIP: 6616@sip.us.mysql.com
Mobile: +61 4 3 8844 332

Jumpstart your cluster:
http://www.mysql.com/consulting/packaged/cluster.html


application/pgp-signature attachment: This is a digitally signed message part