Re: Cluster performance with data node down

On Fri, 2007-08-31 at 16:57 +0200, Giulio Ferro wrote:
> The thing I'm still wondering at is the huge gap of performance for this
> kind of queries
> between a full working configuration (2/2 dn) and degraded cluster (1/2 dn)
>
> (btw condition pushdown seems to be on by default in 5.1.20/21)

yes, although there's some situations where the mysql server does things better... so good to try with it disabled.

> As I stated in my first message, after the first query (which takes
> about 12 secs)
> the following queries, even though they are full scan, take about 0.4
> seconds,
> which is very acceptable in my opinion.

you could have query cache enabled... (although I'd expect it to be even quicker then)

> The true problem is when a data node fails or is brought down. In this
> event the
> time for a query skyrockets to 12 secs, which of course is unacceptable
> in many
> production environment.

Possibly file a bug with a reproducable test case.

> I still fail to see how parallelization in a query like this could
> account for such
> a gap in performance (0.4 s. <-> 12 s.)
> Since the degraded time is the same time it takes for the first
> execution of queries
> I guess that condition pushdown doesn't work for a degraded cluster.

it should.

Do you need help?

In addition to our free technical library, Pantek provides professional IT services to companies who utilize Open Source software: Learn more about Pantek's Technical Support offerings Place an order for 24/7/365 Technical Support here Learn more about Pantek, who we are, and what we do Chat with a Live Operator Related lists.mysql.com Links announce benchmarks bugs certification cluster commits dotnet gui-tools internals java maxdb myodbc mysql mysqldoc packagers perl plusplus replication win32 Request more information about Pantek

> Please don't let this issue fall. Textual 'like' searches are very
> important in many
> contexts (eg. search engines in web sites, name searches in organizer
> applications, etc.)
> If performance stays like, this the usefulness of a cluster
> architecture is considerably
> lessened.

The way to not let something like this fall away is to file a bug in the bugs system. that's where we track such things.

-- 
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