I recently had the chance to play with the new T5140 servers. Using the Sparc CMT architecture, these servers present an amazing 128 cpu's to you to use (as a combination of cores and compute threads, there are 2 sockets only).

We are doing some trials with eager Sun customers who want to utilize these babies. The good news is that MySQL Cluster 7.0 (aka 6.4) will support a multi-threaded data node option. The bad news is, one ndbd process still only uses about 8 CPU cores, so to utilize 128, there is some way to go! So the bad news is we still have to launch many ndbd processes to get out the full power of these boxes. But the good news is that with 7.0 there is at least a point in trying at all.

I developed a simple script which lets me easily start a varying amount of ndbd and mysqld processes on one host (and then copy the script to also start same amount of processes on another host). If you have been using Johan's excellent …

I recently had the chance to play with the new T5140 servers. Using the Sparc CMT architecture, these servers present an amazing 128 cpu's to you to use (as a combination of cores and compute threads, there are 2 sockets only).

We are doing some trials with eager Sun customers who want to utilize these babies. The good news is that MySQL Cluster 7.0 (aka 6.4) will support a multi-threaded data node option. The bad news is, one ndbd process still only uses about 8 CPU cores, so to utilize 128, there is some way to go! So the bad news is we still have to launch many ndbd processes to get out the full power of these boxes. But the good news is that with 7.0 there is at least a point in trying at all.

I developed a simple script which lets me easily start a varying amount of ndbd and mysqld processes on one host (and then copy the script to also start same amount of processes on another host). If you have been using Johan's excellent …

The MySQL Cluster product seems to intimidate a lot of folks with its complexity. The North Texas MySQL Meetup had a presentation at the January meeting titled 'A Gentle Introduction to MySQL Cluster' where a cluster was setup on a single laptop. The presentation was full of information that needs to be shared to encourage others to try out this valuable piece of software.

The following covers how to set up a cluster on a single computer. The cluster will consist of two data nodes, a SQL node, and a management node. In this case the SQL node and the management node will be on one system. The data nodes will be virtual systems running on the same host. This is not the optimal performance configuration but it will let those new to MySQL Cluster try out the product with a minimal investment in equipment and time.

You will need two free software products -- …

One thing we were guessing at with Massimo yesterday is, if you store
large blobs as disk data, how much will they consume DataMemory and
IndexMemory (primary key, each "chunk" has a hidden primary key, first 25x bytes
of blob stored in memory...)?

My empirical test showed that about 2% of the total size of blobs is needed for RAM
(25% of that is IndexMemory).

IMHO this is close to negligible, but in many situations not negligible
at all (may have close to TB of disk data -> 20GB of RAM needed for
disk data).

Also note that this is a minimum figure. If you actually have something
else than the blob (like other indexes) you of course use much more RAM.

The test was:
CREATE TABLE `jpgtest` (
`id` int(11) NOT NULL,
`jpg` blob,
PRIMARY KEY (`id`)
) TABLESPACE ts_1 STORAGE DISK ENGINE=ndbcluster;

One thing we were guessing at with Massimo yesterday is, if you store
large blobs as disk data, how much will they consume DataMemory and
IndexMemory (primary key, each "chunk" has a hidden primary key, first 25x bytes
of blob stored in memory...)?

My empirical test showed that about 2% of the total size of blobs is needed for RAM
(25% of that is IndexMemory).

IMHO this is close to negligible, but in many situations not negligible
at all (may have close to TB of disk data -> 20GB of RAM needed for
disk data).

Also note that this is a minimum figure. If you actually have something
else than the blob (like other indexes) you of course use much more RAM.

The test was:
CREATE TABLE `jpgtest` (
`id` int(11) NOT NULL,
`jpg` blob,
PRIMARY KEY (`id`)
) TABLESPACE ts_1 STORAGE DISK ENGINE=ndbcluster;

One thing we were guessing at with Massimo yesterday is, if you store
large blobs as disk data, how much will they consume DataMemory and
IndexMemory (primary key, each "chunk" has a hidden primary key, first 25x bytes
of blob stored in memory...)?

My empirical test showed that about 2% of the total size of blobs is needed for RAM
(25% of that is IndexMemory).

IMHO this is close to negligible, but in many situations not negligible
at all (may have close to TB of disk data -> 20GB of RAM needed for
disk data).

Also note that this is a minimum figure. If you actually have something
else than the blob (like other indexes) you of course use much more RAM.

The test was:
CREATE TABLE `jpgtest` (
`id` int(11) NOT NULL,
`jpg` blob,
PRIMARY KEY (`id`)
) TABLESPACE ts_1 STORAGE DISK ENGINE=ndbcluster;

Last week I helped a customer setup a JBoss application against MySQL Cluster. It turns out it is not immediately obvious how you should setup our JDBC connector to do loadbalancing and failover. For instance, setting the connector up for an Master-Slave setup (with MySQL Enterprise) is well documented, but not doing the same with MySQL Cluster.

It's not really properly documented in the manual part, but I found in the changelogs, and confirmed on IRC that to do load-balancing across the SQL nodes in MySQL Cluster, you would use a different JDBC connection string with the "loadbalance" keyword added...

jdbc:mysql:loadbalance://host-1,host-2,...host-n/database

That does indeed …

Last week I helped a customer setup a JBoss application against MySQL Cluster. It turns out it is not immediately obvious how you should setup our JDBC connector to do loadbalancing and failover. For instance, setting the connector up for an Master-Slave setup (with MySQL Enterprise) is well documented, but not doing the same with MySQL Cluster.

It's not really properly documented in the manual part, but I found in the changelogs, and confirmed on IRC that to do load-balancing across the SQL nodes in MySQL Cluster, you would use a different JDBC connection string with the "loadbalance" keyword added...

jdbc:mysql:loadbalance://host-1,host-2,...host-n/database

That does indeed …

Last week I helped a customer setup a JBoss application against MySQL Cluster. It turns out it is not immediately obvious how you should setup our JDBC connector to do loadbalancing and failover. For instance, setting the connector up for an Master-Slave setup (with MySQL Enterprise) is well documented, but not doing the same with MySQL Cluster.

It's not really properly documented in the manual part, but I found in the changelogs, and confirmed on IRC that to do load-balancing across the SQL nodes in MySQL Cluster, you would use a different JDBC connection string with the "loadbalance" keyword added...

jdbc:mysql:loadbalance://host-1,host-2,...host-n/database

That does indeed …

I had a very interesting conversation on the phone with Jonas
Oreland today (he also blogged about it on his blog at
http://jonasoreland.blogspot.com).

There is a lot of interesting features coming up in MySQL Cluster
version 6.4. Online Add Node is one of those, which can be done
without any downtime and even with almost no additional memory
needed other than the memory in the new machines added into the
cluster. This is a feature I started thinking almost 10 years ago
so it's nice to see the fourth version of the solution actually be
implemented and it's a really neat solution to the problem,
definitely fitting the word innovative.

The next interesting feature is to use a more efficient protocol
for handling large operations towards the data nodes. This makes it
use less bits …