Somewhat obscure Q about clustering

mairhtin o'feannag

Hello,

I have what may seem a bit of an odd question, but here goes:

Is there any advantage to having an operating-system clustered set of
machines running DB2, as opposed to using DB2's PDF to "cluster" machines?
The discussion came up due to a cluster of Linux servers we have which are
not being used very much. The suggestion was to use them as a database
cluster. I actually had no idea on how to answer this question, since I
would imagine that even in an OS cluster, you would install DB2 on each
machine and make a node of each machine?

*scratching head as to what the implications are*

Mairhtin

Nov 12 '05 #1

Haider Rizvi

Mairhtin,

I am not sure of the distinction that you are making, can you elaborate
a bit more on what you mean by operating-system clustered set of
machines? Are you thinking of numa-type machines providing a single
system image from an OS perspective?

Whether it is a large SMP server (a la IBM p690, Sun starfire), a NUMA
box (e.g. from IBM x445, SGI) or a cluster of small nodes (cluster of
e325 servers, p570s, etc.), DB2 DPF maps pretty well on either
architectures. We have published industry-std benchmarks as well as
have a large number of customers using these various architectures in
their environments.

- Haider

"mairhtin o'feannag" <ir**********@rocketmaildot.com> writes:

Hello,

I have what may seem a bit of an odd question, but here goes:

Is there any advantage to having an operating-system clustered set of machines
running DB2, as opposed to using DB2's PDF to "cluster" machines? The
discussion came up due to a cluster of Linux servers we have which are not being
used very much. The suggestion was to use them as a database cluster. I
actually had no idea on how to answer this question, since I would imagine that
even in an OS cluster, you would install DB2 on each machine and make a node of
each machine?

*Scratching head as to what the implications are*

Mairhtin

Nov 12 '05 #2

Mairhtin O'Feannag

I'm talking about 10 machines that are a linux-cluster. Intel machines
running clustering software. ROCKS, I believe, but the same question
would exist for a Beowulf Cluster as well. The machines all have their
own hard drives.

Mairhtin


Haider Rizvi <ha*****@gmail.com> wrote in
news:7z************@thinkhr.torolab.ibm.com:

Mairhtin,

I am not sure of the distinction that you are making, can you
elaborate a bit more on what you mean by operating-system clustered
set of machines? Are you thinking of numa-type machines providing a
single system image from an OS perspective?

Whether it is a large SMP server (a la IBM p690, Sun starfire), a NUMA
box (e.g. from IBM x445, SGI) or a cluster of small nodes (cluster of
e325 servers, p570s, etc.), DB2 DPF maps pretty well on either
architectures. We have published industry-std benchmarks as well as
have a large number of customers using these various architectures in
their environments.

- Haider

"mairhtin o'feannag" <ir**********@rocketmaildot.com> writes:

Hello,

I have what may seem a bit of an odd question, but here goes:

Is there any advantage to having an operating-system clustered set of
machines running DB2, as opposed to using DB2's PDF to "cluster"
machines? The discussion came up due to a cluster of Linux servers
we have which are not being used very much. The suggestion was to
use them as a database cluster. I actually had no idea on how to
answer this question, since I would imagine that even in an OS
cluster, you would install DB2 on each machine and make a node of
each machine?

*Scratching head as to what the implications are*

Mairhtin

Nov 12 '05 #3

Ian

Mairhtin O'Feannag wrote:

I'm talking about 10 machines that are a linux-cluster. Intel machines
running clustering software. ROCKS, I believe, but the same question
would exist for a Beowulf Cluster as well. The machines all have their
own hard drives.

Beowulf is a set of tools used to allow a program to run on a group
("cluster") of machines -- it takes care of communication between
the nodes, distributing work to the nodes, etc.

The Database Partitioning Feature serves the equivalent purpose, but
it's built into DB2 UDB. You can't replace DPF with Beowulf.

Also keep in mind that using a cluster to scale out (using DPF or
Beowulf, if you're building a compute-farm) is not the same thing as
scaling for high availability (i.e. using a product like Veritas Cluster
or Steeleye Lifekeeper).
----== Posted via Newsfeeds.Com - Unlimited-Uncensored-Secure Usenet News==----
http://www.newsfeeds.com The #1 Newsgroup Service in the World! >100,000 Newsgroups
---= East/West-Coast Server Farms - Total Privacy via Encryption =---

Nov 12 '05 #4

Haider Rizvi

"Mairhtin O'Feannag" <ir********@rocketmail.com> writes:

I'm talking about 10 machines that are a linux-cluster. Intel machines running
clustering software. Rocks, I believe, but the same question would exist for a
Beowulf Cluster as well. The machines all have their own hard drives.
I was confused by the way you posed your question.

DB2 + DPF is a shared-nothing parallel database that can run on a
cluster of nodes. Other than the basic linux OS (typically RHEL3 or
SLES8 these days for enterprise deployment) you don't need any special
software on these nodes (which is what Beowulf seems to be providing, I
don't know much about Beowulf). You can use a variety of software to
maintain / sys admin such a cluster, examples of such software are IBM
CSM, Power Cockpit, xCAT, etc.

You can read more details about dpf in the DB2 books, or various
articles at developerworks, etc. One overview article is at the
following url:

http://www-106.ibm.com/developerwork.../dm-0403chong/

See specific db2 for linux clusters papers and related information at:
http://www-306.ibm.com/software/data...ux/papers.html
Haider Rizvi <ha*****@gmail.com> wrote in
news:7z************@thinkhr.torolab.ibm.com:

Mairhtin,

I am not sure of the distinction that you are making, can you elaborate a bit
more on what you mean by operating-system clustered set of machines? Are you
thinking of numa-type machines providing a single system image from an OS
perspective?

Whether it is a large SMP server (a la IBM p690, Sun starfire), a NUMA box
(e.g. from IBM x445, SGI) or a cluster of small nodes (cluster of e325
servers, p570s, etc.), DB2 DPF maps pretty well on either architectures. We
have published industry-std benchmarks as well as have a large number of
customers using these various architectures in their environments.

- Haider

"mairhtin o'feannag" <ir**********@rocketmaildot.com> writes:

Hello,

I have what may seem a bit of an odd question, but here goes:

Is there any advantage to having an operating-system clustered set of
machines running DB2, as opposed to using DB2's PDF to "cluster" machines?
The discussion came up due to a cluster of Linux servers we have which are
not being used very much. The suggestion was to use them as a database
cluster. I actually had no idea on how to answer this question, since I
would imagine that even in an OS cluster, you would install DB2 on each
machine and make a node of each machine?

*Scratching head as to what the implications are*

Mairhtin

--
Regards,
--
Haider

Nov 12 '05 #5

Mairhtin O'Feannag

Thank you all.

That's pretty much what I expected, but needed to be certain I wasn't
blowing smoke.... :)

So on with the DPF show!

Mairhtin
Haider Rizvi <ha*****@gmail.com> wrote in
news:7z************@thinkhr.torolab.ibm.com:

"Mairhtin O'Feannag" <ir********@rocketmail.com> writes:

I'm talking about 10 machines that are a linux-cluster. Intel
machines running clustering software. Rocks, I believe, but the same
question would exist for a Beowulf Cluster as well. The machines all
have their own hard drives.

I was confused by the way you posed your question.

DB2 + DPF is a shared-nothing parallel database that can run on a
cluster of nodes. Other than the basic linux OS (typically RHEL3 or
SLES8 these days for enterprise deployment) you don't need any special
software on these nodes (which is what Beowulf seems to be providing,
I don't know much about Beowulf). You can use a variety of software to
maintain / sys admin such a cluster, examples of such software are IBM
CSM, Power Cockpit, xCAT, etc.

You can read more details about dpf in the DB2 books, or various
articles at developerworks, etc. One overview article is at the
following url:

http://www-106.ibm.com/developerwork...icle/dm-0403ch
ong/

See specific db2 for linux clusters papers and related information at:
http://www-306.ibm.com/software/data...ux/papers.html

Haider Rizvi <ha*****@gmail.com> wrote in
news:7z************@thinkhr.torolab.ibm.com:

Mairhtin,

I am not sure of the distinction that you are making, can you
elaborate a bit more on what you mean by operating-system clustered
set of machines? Are you thinking of numa-type machines providing a
single system image from an OS perspective?

Whether it is a large SMP server (a la IBM p690, Sun starfire), a
NUMA box (e.g. from IBM x445, SGI) or a cluster of small nodes
(cluster of e325 servers, p570s, etc.), DB2 DPF maps pretty well on
either architectures. We have published industry-std benchmarks as
well as have a large number of customers using these various
architectures in their environments.

- Haider

"mairhtin o'feannag" <ir**********@rocketmaildot.com> writes:

Hello,

I have what may seem a bit of an odd question, but here goes:

Is there any advantage to having an operating-system clustered set
of machines running DB2, as opposed to using DB2's PDF to "cluster"
machines? The discussion came up due to a cluster of Linux servers
we have which are not being used very much. The suggestion was to
use them as a database cluster. I actually had no idea on how to
answer this question, since I would imagine that even in an OS
cluster, you would install DB2 on each machine and make a node of
each machine?

*Scratching head as to what the implications are*

Mairhtin

Nov 12 '05 #6