Hi,
Brian Z wrote:
> Hello list,
>
> I have been following with great interest the
> developments in Mysql Cluster. If someone could
> point me in the right direction for answers or
> give some insight, I would appreciate it.
>
> Is there a roadmap for release dates, specifically
> for the 5.1 GA release.
>
We are in alpha phase and will hit the beta phase hopefully in not too
long time, from there it's pretty much a stabilsation period of 5.1.
> Is there anywhere I can follow development on the
> hot plug in/out of servers into Cluster? Backing
> up, plugging in a new storage server, and
> restoring is not acceptable in our environment
> just to grow capacity.
>
Actually yes, MySQL delivers all its commit emails to a special
email list. So it is possible to follow all the development on this
topic and on all other topics as well.
2 weeks ago I saw the first adding of a node to a running cluster, it
was awesome to see it after working with this limitation for such
a long time.
If you search for WL 1504 in the commit logs you'll find most of the
commits related to this area. This development is about being able
to add/drop nodes and/or node groups to a running cluster.
There is another WL 2498 that deals with reorganising the data already
in the cluster. This is a very interesting development and it is among the
more complex problems I came across so far in the cluster development.
It's complexity comes from the fact that it involves 3 phases of 14 steps
and one has to take into account node crashes and system crashes in all
those stages. So to handle this complexity we've come up with a schema
operation scheme such that we later on will be able to construct very complex
change operations of schemas, nodes, replicas and so forth without a very
big development.
> Is there somewhere I can find detailed
> specifications on disk-based NDB? I need to know
> planned database size limitations, and overhead
> involved in management of very large datasets.
>
Disk-data hasn't yet been pushed to 5.1, as soon as this happens the
documentation will arrive. The major limitation to note is that hash
indexes and indexed fields and indexes still has to be in memory.
Thus it depends a bit on your data what can be put into the cluster.
It will be a major improvement compared to 4.1 and 5.0 but it isn't
yet possible to store any data in the cluster.
Rgrds Mikael
> Our datasets range from a few hundred megabytes,
> to about 1.3TB with about 10TB data total in our
> active databases. There's another 1TB of nearline
> log data which requires extremely high write rates
> from a cluster of 60 realtime application servers,
> and high read requirements when reports are run
> off the logs.
>
> We have 4 quad Xeon 3.66Ghz 16GB ram 2x36GB 10k
> rpm 6x146GB 15k rpm drives. I need to start
> migrating some data into a MySQL Cluster due to
> another db's licencing costs and our old MySQL
> installations' performance, but the above are all
> show stoppers until they can be dealt with.
>
> Brian
>
>
>
> Edited 2 times. Last edit at 12/15/05 09:34PM by
> Brian Z.
Mikael Ronstrom
Senior Software Architect, MySQL AB
My blog:
http://mikaelronstrom.blogspot.com