Thursday, March 22, 2012

ACTIVE/ACTIVE configuration

We are looking at installing a 2 Node ACTIVE/ACTIVE SQL
Server configuration utilizing two DELL PE6600 servers,
both connected via FIBRE to a DELL CX200(CLARION). The
OS will be clustered in a ACTIVE/ACTIVE configuration as
well.
The main purpose for attempting this is more for
spreading processing across systems more than for
failover capabilities.
I have a few questions, the first would be has anyone
accomplished this? Secondly, in a SQL server
ACTIVE/ACTIVE configuration, can both instances utilize
one shared disk configuration, so that database merging
does not need to happen?
Any help/advice would be greatly appreciated as i'm
drawing up a plan.
Failover clustering is designed for high availability. Failover clustering
is NOT a scale-out technology. There are some scenarios where data
partitioning and merge replication can be used in conjunction with
multi-instance clustering to achieve scale-out, but they require intensive
design and management. You are probably better off buying a bigger box for
your server than trying to scale out.
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Jeff" <anonymous@.discussions.microsoft.com> wrote in message
news:81e201c431f4$708e9430$a501280a@.phx.gbl...
> We are looking at installing a 2 Node ACTIVE/ACTIVE SQL
> Server configuration utilizing two DELL PE6600 servers,
> both connected via FIBRE to a DELL CX200(CLARION). The
> OS will be clustered in a ACTIVE/ACTIVE configuration as
> well.
> The main purpose for attempting this is more for
> spreading processing across systems more than for
> failover capabilities.
>

> I have a few questions, the first would be has anyone
> accomplished this? Secondly, in a SQL server
> ACTIVE/ACTIVE configuration, can both instances utilize
> one shared disk configuration, so that database merging
> does not need to happen?
>
No.

> Any help/advice would be greatly appreciated as i'm
> drawing up a plan.
|||I guess with that being said, would you have any ideas in
regards to distributed processing.
High availiability is what i'm looking for, just in an
active/active configuration. Can I not have a sql server
running on two different systems and look at the same
logical data source via SAN solution?

>--Original Message--
>Failover clustering is designed for high availability.
Failover clustering
>is NOT a scale-out technology. There are some scenarios
where data
>partitioning and merge replication can be used in
conjunction with
>multi-instance clustering to achieve scale-out, but they
require intensive
>design and management. You are probably better off
buying a bigger box for
>your server than trying to scale out.
>--
>Geoff N. Hiten
>Microsoft SQL Server MVP
>Senior Database Administrator
>Careerbuilder.com
>I support the Professional Association for SQL Server
>www.sqlpass.org
>"Jeff" <anonymous@.discussions.microsoft.com> wrote in
message[vbcol=seagreen]
>news:81e201c431f4$708e9430$a501280a@.phx.gbl...
as
>No.
>
>.
>
|||SQL Failover clustering works on a shared-nothing setup. Each instance has
exclusive control over the underlying resources, including its portion of
the shared storage system. The cluster service arbitrates ownership of the
underlying resources so that only one host node at a time controls each
resource.
Distributed database computing is not quite what is advertised. The TPC-C
benchmark is particularly well suited to a distributed model (coincidence
maybe?), but not all applications work as well. Buying a larger box is
probably where you want to go if you need more processing power.
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Jeff" <anonymous@.discussions.microsoft.com> wrote in message
news:830d01c43207$a14ec370$a601280a@.phx.gbl...[vbcol=seagreen]
> I guess with that being said, would you have any ideas in
> regards to distributed processing.
> High availiability is what i'm looking for, just in an
> active/active configuration. Can I not have a sql server
> running on two different systems and look at the same
> logical data source via SAN solution?
>
> Failover clustering
> where data
> conjunction with
> require intensive
> buying a bigger box for
> message
> as
|||That makes since. I have been doing some reading on
active/active cluster configurations. Just want to
understand that in that configuration each instance
serves it's own needs until the time that one fails and
then the other instance picks up the additional workload
of the failed instance. Is that correct?

>--Original Message--
>SQL Failover clustering works on a shared-nothing
setup. Each instance has
>exclusive control over the underlying resources,
including its portion of
>the shared storage system. The cluster service
arbitrates ownership of the
>underlying resources so that only one host node at a
time controls each
>resource.
>Distributed database computing is not quite what is
advertised. The TPC-C
>benchmark is particularly well suited to a distributed
model (coincidence
>maybe?), but not all applications work as well. Buying
a larger box is
>probably where you want to go if you need more
processing power.
>--
>Geoff N. Hiten
>Microsoft SQL Server MVP
>Senior Database Administrator
>Careerbuilder.com
>I support the Professional Association for SQL Server
>www.sqlpass.org
>"Jeff" <anonymous@.discussions.microsoft.com> wrote in
message[vbcol=seagreen]
>news:830d01c43207$a14ec370$a601280a@.phx.gbl...
in[vbcol=seagreen]
server[vbcol=seagreen]
scenarios[vbcol=seagreen]
they[vbcol=seagreen]
SQL[vbcol=seagreen]
servers,[vbcol=seagreen]
The[vbcol=seagreen]
configuration[vbcol=seagreen]
anyone[vbcol=seagreen]
utilize[vbcol=seagreen]
merging
>
>.
>
|||Nope. You are confusing instances and hosts. Not surprising since
active-active is a holdover from the SQL 7.0 days and is misleading for SQL
2000.
Lets start with a single-instance cluster. Two nodes (host computers), one
SQL instance. The instance runs on one host at a time. The cluster
services on all nodes monitor the running instance and if it fails a series
of tests (looks-alive and is-alive) the instance is deemed failed and is
restarted. The restart may be on the original node or on the other node.
You can add a second, independant instance that normally runs on the second
node but can run on either node. The two instances have nothing in common
except the set of host nodes that they can run on. There are some extra
concerns when using multiple instances on a cluster to make sure they can
'stack' on a single host node if necessary, but otherwise they do not
interact in any way. People incorrectly refer to this as 'Active-Active',
meaning both nodes have active instances. However, under SQL 2000 you can
have up to 16 instances on a cluster, so the naming quickly can get out of
hand.
Under SQL 7.0 the nodes were not completely interchangable so the
active-passive and active-active names were important. Each node owned a
single instance and the other node could run it but not make certain changes
to the setup. Now, single-instance and multiple-instance naming is more
meaningful since the nodes are truly interchangable. This is especially
true now that Windows 2003 allows 8 node clusters using Enterprise Edition.
(Active-Active-Passive-Passive-Active.. Blah, Blah, Blah).
Sounds like you need a single-instance installation running on fairly beefy
hardware. FYI, the CX-200 has been replaced with the CX-300. If you go
with a system larger than a 6600/6650 I would strongly look at the CX-500 or
CX-700 for the performance benefits.
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Jeff" <anonymous@.discussions.microsoft.com> wrote in message
news:837001c43213$70e59cc0$a501280a@.phx.gbl...[vbcol=seagreen]
> That makes since. I have been doing some reading on
> active/active cluster configurations. Just want to
> understand that in that configuration each instance
> serves it's own needs until the time that one fails and
> then the other instance picks up the additional workload
> of the failed instance. Is that correct?
> setup. Each instance has
> including its portion of
> arbitrates ownership of the
> time controls each
> advertised. The TPC-C
> model (coincidence
> a larger box is
> processing power.
> message
> in
> server
> scenarios
> they
> SQL
> servers,
> The
> configuration
> anyone
> utilize
> merging
|||No. SQL Server does not allow shared access to a database. A single SQL
Server can open the database files and access them and that prevents
anything else for getting at the data. This is a data protection mechanism
and exists to prevent possible corruption of the data when two independent
resources attempt to write the same data.
Mike
Principal Mentor
Solid Quality Learning
"More than just Training"
SQL Server MVP
http://www.solidqualitylearning.com
http://www.mssqlserver.com
|||You may deploy it on SAN
|||I was going to have my two nodes on a SAN solution. Is
there differences from what the other gentleman was
stating in regards to having a SAN solution?

>--Original Message--
>You may deploy it on SAN
>.
>
|||You can store SQL data on a SAN. It is generally the best way to build a
cluster.
Geoff N. Hiten
Microsoft SQL Server MVP
Senior Database Administrator
Careerbuilder.com
I support the Professional Association for SQL Server
www.sqlpass.org
"Jeff" <anonymous@.discussions.microsoft.com> wrote in message
news:acbc01c4369c$fd6a3290$a601280a@.phx.gbl...[vbcol=seagreen]
> I was going to have my two nodes on a SAN solution. Is
> there differences from what the other gentleman was
> stating in regards to having a SAN solution?

No comments:

Post a Comment