 |
MySQLTalk.org MYSQL discussions groups
|
| View previous topic :: View next topic |
| Author |
Message |
christian johansson Guest
|
Posted: Fri Mar 30, 2007 1:18 am Post subject: only identical machines in a cluster? |
|
|
In the mysql cluster concept basics on mysql.com, I read:
"Configuration of a cluster involves configuring each individual node in the
cluster and setting up individual communication links between nodes. MySQL
Cluster is currently designed with the intention that data nodes are
homogeneous in terms of processor power, memory space, and bandwidth. In
addition, to provide a single point of configuration, all configuration data
for the cluster as a whole is located in one configuration file."
If I don't misunderstand this completely, this means that in order to setup
a mysql cluster, all machines in it have to be identical to each other. Is
that really the case?
What kind of performance penalties do you receive if you use un-equal
machines? Does that make them all perform only so fast as the weakest link
then?
Thanks
Christian |
|
| Back to top |
|
 |
Geert Vanderkelen Guest
|
Posted: Fri Mar 30, 2007 1:27 am Post subject: Re: only identical machines in a cluster? |
|
|
Hey Christian,
On 29 Mar 2007, at 22:18, christian johansson wrote:
| Quote: | In the mysql cluster concept basics on mysql.com, I read:
"Configuration of a cluster involves configuring each individual
node in the
cluster and setting up individual communication links between
nodes. MySQL
Cluster is currently designed with the intention that data nodes are
homogeneous in terms of processor power, memory space, and
bandwidth. In
addition, to provide a single point of configuration, all
configuration data
for the cluster as a whole is located in one configuration file."
If I don't misunderstand this completely, this means that in order
to setup
a mysql cluster, all machines in it have to be identical to each
other. Is
that really the case?
|
Data nodes don't have to be identical to the MySQL server using them.
The SQL nodes can run on less or more power full machines.
| Quote: | What kind of performance penalties do you receive if you use un-equal
machines? Does that make them all perform only so fast as the
weakest link
then?
|
It's not about performance, it's about availability. Data nodes are
best to be equal because one takes over its partner when it gets
down. Also, if you need to replace one of them, you know exactly what
you need. In most cases, you want to have a spare data node which you
can put immediately in with the others. And there for it is good to
have the Data Nodes equal.
Best regards,
Geert
--
Geert Vanderkelen, Support Engineer
MySQL GmbH, Germany, www.mysql.com
Are you MySQL certified? www.mysql.com/certification |
|
| Back to top |
|
 |
Andrew Hutchings Guest
|
Posted: Fri Mar 30, 2007 10:10 am Post subject: Re: only identical machines in a cluster? |
|
|
Geert Vanderkelen wrote:
| Quote: | Data nodes don't have to be identical to the MySQL server using them.
The SQL nodes can run on less or more power full machines.
|
Although if you mix processor types in a cluster (such as PPC for 1 data
node and x86 for another) you will run into problems.
Regards
Andrew
--
Andrew Hutchings - Linux Jedi - http://www.linuxjedi.co.uk/
"There are two major products that come out of Berkeley: LSD and UNIX.
We don't believe this to be a coincidence." - Jeremy S. Anderson |
|
| Back to top |
|
 |
Jason Brooke Guest
|
Posted: Fri Mar 30, 2007 10:10 am Post subject: Re: only identical machines in a cluster? |
|
|
Would dual core Xeon's (5150) be ok to mix with amd opteron's (252)
Jason
Andrew Hutchings wrote:
| Quote: | Geert Vanderkelen wrote:
Data nodes don't have to be identical to the MySQL server using them.
The SQL nodes can run on less or more power full machines.
Although if you mix processor types in a cluster (such as PPC for 1 data
node and x86 for another) you will run into problems.
Regards
Andrew |
|
|
| Back to top |
|
 |
Geert Vanderkelen Guest
|
Posted: Fri Mar 30, 2007 10:10 am Post subject: Re: only identical machines in a cluster? |
|
|
On Mar 30, 2007, at 08:42, Jason Brooke wrote:
| Quote: | Would dual core Xeon's (5150) be ok to mix with amd opteron's (252)
|
Yes. They are both x86. The endian has to be the same. Like Andrew
said: mixing PowerPC Apples machines with Intel/AMD powered pc's will
not work..
Maybe in the future, who knows :)
| Quote: |
Andrew Hutchings wrote:
Geert Vanderkelen wrote:
Data nodes don't have to be identical to the MySQL server using
them.
The SQL nodes can run on less or more power full machines.
Although if you mix processor types in a cluster (such as PPC for
1 data
node and x86 for another) you will run into problems.
Regards
Andrew
|
--
Geert Vanderkelen, Support Engineer
MySQL GmbH, Germany, www.mysql.com |
|
| Back to top |
|
 |
Stewart Smith Guest
|
Posted: Wed Apr 04, 2007 4:47 pm Post subject: Re: only identical machines in a cluster? |
|
|
On Fri, 2007-03-30 at 07:18 +0100, Andrew Hutchings wrote:
| Quote: | Geert Vanderkelen wrote:
Data nodes don't have to be identical to the MySQL server using them.
The SQL nodes can run on less or more power full machines.
Although if you mix processor types in a cluster (such as PPC for 1 data
node and x86 for another) you will run into problems.
|
or rather a friendly error message
(that was the first patch I added to Cluster :)
There's currently ongoing work to allow mixed endian clusters (focus on
same endian ndbd but mixed api nodes). Don't hold breath... but it's
coming.
--
Stewart Smith, Software Engineer
MySQL AB, www.mysql.com
Office: +14082136540 Ext: 6616
VoIP: 6616 (AT) sip (DOT) us.mysql.com
Mobile: +61 4 3 8844 332
Jumpstart your cluster:
http://www.mysql.com/consulting/packaged/cluster.html |
|
| Back to top |
|
 |
Andrew Hutchings Guest
|
Posted: Thu Apr 05, 2007 1:54 am Post subject: Re: only identical machines in a cluster? |
|
|
Stewart Smith wrote:
| Quote: | or rather a friendly error message
(that was the first patch I added to Cluster :)
There's currently ongoing work to allow mixed endian clusters (focus on
same endian ndbd but mixed api nodes). Don't hold breath... but it's
coming.
|
Surely there can't be many people requiring this nowadays? I would
imagine most people using NDB will be using a bunch of almost identical
hardware, if anything for ease of maintenance.
I am sitting patiently waiting for MySQL 5.1 to go GA anyway because it
would be nice to change a table schema (ie. TRUNCATE) without having to
restart the other MySQL nodes.
Regards
Andrew
--
Andrew Hutchings - Linux Jedi - http://www.linuxjedi.co.uk/
"There are two major products that come out of Berkeley: LSD and UNIX.
We don't believe this to be a coincidence." - Jeremy S. Anderson |
|
| Back to top |
|
 |
Jon Stephens Guest
|
Posted: Thu Apr 05, 2007 5:09 pm Post subject: Re: only identical machines in a cluster? |
|
|
Andrew Hutchings wrote:
| Quote: | Stewart Smith wrote:
or rather a friendly error message
(that was the first patch I added to Cluster :)
There's currently ongoing work to allow mixed endian clusters (focus on
same endian ndbd but mixed api nodes). Don't hold breath... but it's
coming.
Surely there can't be many people requiring this nowadays? I would
imagine most people using NDB will be using a bunch of almost identical
hardware, if anything for ease of maintenance.
I am sitting patiently waiting for MySQL 5.1 to go GA anyway because it
would be nice to change a table schema (ie. TRUNCATE) without having to
restart the other MySQL nodes.
|
You're sure about this? AFAIK, it should just require a "dummy" query
against the table afterwards from each SQL node.
--
Jon Stephens - jon (AT) mysql (DOT) com
Technical Writer - MySQL Documentation Team
Brisbane, Australia (GMT +10.00)
Office: +61 (7) 3209 1394
Mobile: +61 402 635 784
MySQL AB: www.mysql.com |
|
| Back to top |
|
 |
Geert Vanderkelen Guest
|
Posted: Thu Apr 05, 2007 5:15 pm Post subject: Re: only identical machines in a cluster? |
|
|
On Apr 5, 2007, at 14:09, Jon Stephens wrote:
| Quote: | Andrew Hutchings wrote:
Stewart Smith wrote:
or rather a friendly error message
(that was the first patch I added to Cluster :)
There's currently ongoing work to allow mixed endian clusters
(focus on
same endian ndbd but mixed api nodes). Don't hold breath... but it's
coming.
Surely there can't be many people requiring this nowadays? I would
imagine most people using NDB will be using a bunch of almost
identical
hardware, if anything for ease of maintenance.
|
I for one would welcome it. Having my PowerBook and Linux x86
machine, I could save some power on me laptop!
There are also situation where people trust more big Sun machines,
but have all there clients on cheaper x86 hardware for example.
| Quote: | I am sitting patiently waiting for MySQL 5.1 to go GA anyway
because it
would be nice to change a table schema (ie. TRUNCATE) without
having to
restart the other MySQL nodes.
You're sure about this? AFAIK, it should just require a "dummy"
query against the table afterwards from each SQL node.
|
Yes, no need to restart the mysqld. Not nice with the dummy, but it
works.
--
Geert Vanderkelen, Support Engineer
MySQL GmbH, Germany, www.mysql.com |
|
| Back to top |
|
 |
Andrew Hutchings Guest
|
Posted: Thu Apr 05, 2007 6:19 pm Post subject: Re: only identical machines in a cluster? |
|
|
Jon Stephens wrote:
| Quote: | You're sure about this? AFAIK, it should just require a "dummy" query
against the table afterwards from each SQL node.
|
Yea, but its a PITA when you need to make a change to a table structure
on a live system, or a developer runs TRUNCATE without telling you, its
quicker just to do a 'service mysql reload' on the other SQL nodes.
Running a "dummy" query via. PHP doesn't do it unfortunately, otherwise
it wouldn't worry so much.
We once had a situation where this wouldn't work at all, it said some
unknown error. The only way around it was to drop the table and
re-create it, even a full SQL node restart wouldn't fix it. If it
happens again I will make sure I am told about it so I can file a bug
report.
Regards
Andrew
--
Andrew Hutchings - Linux Jedi - http://www.linuxjedi.co.uk/
"There are two major products that come out of Berkeley: LSD and UNIX.
We don't believe this to be a coincidence." - Jeremy S. Anderson |
|
| Back to top |
|
 |
Kevin Clark Guest
|
Posted: Fri Apr 27, 2007 9:07 pm Post subject: Re: only identical machines in a cluster? |
|
|
Hi,
I'd like to pick up on the point made about using NDB on "almost identical"
hardware - http://lists.mysql.com/cluster/4900
My own experience is that that in three years time the inexpensive commodity
hardware I purchased will be obsolete and the only affordable option will be
to buy the current CPU/RAM combination which performs twice as fast as that
of my 'old' servers.
My question is what impact will adding these new more powerful data nodes
have upon the cluster? Can it cope with having half the data nodes perform
twice as fast the other?
I'd assumed the situation would be analogous to using memory DIMMs of
different frequencies on the same motherboard - whereby memory performance
is limited to that of the frequency of the 'slowest' DIMM - but I'm
struggling to find on the WWW any practical guidance or experiences of
scaling-out a MySQL Cluster further down the line.
Any pointers as to how you managed or are planning to manage scaling-out
your MySQL cluster would be gratefully received.
Many thanks,
Kevin
Connection Software
www.csoft.co.uk |
|
| Back to top |
|
 |
Stewart Smith Guest
|
Posted: Tue May 08, 2007 9:05 am Post subject: Re: only identical machines in a cluster? |
|
|
On Fri, 2007-04-27 at 17:07 +0100, Kevin Clark wrote:
| Quote: | My question is what impact will adding these new more powerful data nodes
have upon the cluster? Can it cope with having half the data nodes perform
twice as fast the other?
I'd assumed the situation would be analogous to using memory DIMMs of
different frequencies on the same motherboard - whereby memory performance
is limited to that of the frequency of the 'slowest' DIMM - but I'm
struggling to find on the WWW any practical guidance or experiences of
scaling-out a MySQL Cluster further down the line.
|
Things will cope fine - and if you just use data on the faster machine,
it'll be faster than before!
Of course, "global" performance is going to be limited by the slowest
point.
--
Stewart Smith, Software Engineer
MySQL AB, www.mysql.com
Office: +14082136540 Ext: 6616
VoIP: 6616 (AT) sip (DOT) us.mysql.com
Mobile: +61 4 3 8844 332
Jumpstart your cluster:
http://www.mysql.com/consulting/packaged/cluster.html |
|
| Back to top |
|
 |
|
|
You cannot post new topics in this forum You cannot reply to topics in this forum You cannot edit your posts in this forum You cannot delete your posts in this forum You cannot vote in polls in this forum
|
|