Showing posts with label clustering. Show all posts
Showing posts with label clustering. Show all posts

Wednesday, March 21, 2012

how many number of nodes

Hi,
I am a newbie to Clustering. I dont know how to decide how many number
of nodes I need to install. For example, if I have 4 SQL database servers
and all 4 servers should be highly availabe and automatically failover. In
this case how can I configure clustering?
Up to 5 or 6. If you want one instance per server - due to load, then 4 is
the min. 1 or 2 boxes for failover, depending on your needs.
Cheers,
Rodney R. Fournier
MVP - Windows Server - Clustering
http://www.nw-america.com - Clustering Website
http://www.msmvps.com/clustering - Blog
http://www.clusterhelp.com - Cluster Training
ClusterHelp.com is a Microsoft Certified Gold Partner
"Bharathi" <NoSpam_BB@.hotmail.com> wrote in message
news:OWOGrVtXHHA.3952@.TK2MSFTNGP04.phx.gbl...
> Hi,
> I am a newbie to Clustering. I dont know how to decide how many
> number of nodes I need to install. For example, if I have 4 SQL database
> servers and all 4 servers should be highly availabe and automatically
> failover. In this case how can I configure clustering?
>
>
|||There are two general configurations: N + I and failover pairs.
For N + I, you would have N active nodes and I inactive. Each of the active
nodes would be paired with one or more of the inactive nodes. If you had
multiple active node failures then the inactive node(s) would have to have
enough processing power to host each of the failed configurations.
For the failover pairs, say you had, in your case, servers A, B, C, and D.
Then you would install each of the instances to a pair: Instance 1 to A and
B, Instance 2 to B and C, Instance 3 to C and D, and Instance 4 to D and A.
All would be active, each capable of running at least 2 instances for
maintenance or due to failure.
I prefer the N + I model because it is easier to manage, but the other works
well too.
Sincerely,
Anthony Thomas

"Bharathi" <NoSpam_BB@.hotmail.com> wrote in message
news:OWOGrVtXHHA.3952@.TK2MSFTNGP04.phx.gbl...
> Hi,
> I am a newbie to Clustering. I dont know how to decide how many
number
> of nodes I need to install. For example, if I have 4 SQL database servers
> and all 4 servers should be highly availabe and automatically failover. In
> this case how can I configure clustering?
>
>
|||For N+I, N+1 is a common practice with a single node acting as the failover
for all the N active nodes. In this case, you basically design to protect
again a single node failure and consider two or more nodes failed at the same
time as not in scope or as a low probability scenario in which multiple
instances will be running on the same failover node with degraded
performance. The main advanatge of N+1 (as opposed to multiple pairs of
tw-node clusters) is the cost saving in hardware (i.e. fewer servers).
We don't configure N+2 or N+3 and so on. But somebody else may chime to
share their experience on having two or more spare nodes.
Linchi
"Anthony Thomas" wrote:

> There are two general configurations: N + I and failover pairs.
> For N + I, you would have N active nodes and I inactive. Each of the active
> nodes would be paired with one or more of the inactive nodes. If you had
> multiple active node failures then the inactive node(s) would have to have
> enough processing power to host each of the failed configurations.
> For the failover pairs, say you had, in your case, servers A, B, C, and D.
> Then you would install each of the instances to a pair: Instance 1 to A and
> B, Instance 2 to B and C, Instance 3 to C and D, and Instance 4 to D and A.
> All would be active, each capable of running at least 2 instances for
> maintenance or due to failure.
> I prefer the N + I model because it is easier to manage, but the other works
> well too.
> Sincerely,
>
> Anthony Thomas
>
> --
> "Bharathi" <NoSpam_BB@.hotmail.com> wrote in message
> news:OWOGrVtXHHA.3952@.TK2MSFTNGP04.phx.gbl...
> number
>
>
|||I need to setup up N+1 cluster.
We have 4 win2003 Enterprise servers: A, B, C, D. I need to install one
instance of sql2000 onto each machine except D which will be the passive
node. My questions:
1. What are the steps to setup a 4-node cluster? (I have done 2-node
cluster.)
2. If each active node uses 4 letters for shared drives, the total number of
active nodes is limited to < = 6, correct?
Thanks a lot.
Jin
"Linchi Shea" wrote:
[vbcol=seagreen]
> For N+I, N+1 is a common practice with a single node acting as the failover
> for all the N active nodes. In this case, you basically design to protect
> again a single node failure and consider two or more nodes failed at the same
> time as not in scope or as a low probability scenario in which multiple
> instances will be running on the same failover node with degraded
> performance. The main advanatge of N+1 (as opposed to multiple pairs of
> tw-node clusters) is the cost saving in hardware (i.e. fewer servers).
> We don't configure N+2 or N+3 and so on. But somebody else may chime to
> share their experience on having two or more spare nodes.
> Linchi
> "Anthony Thomas" wrote:

How many networks does clustering need ?

I know there is a public network for all application communication and one
for hearbeat .
Are there any more ?
Does one connect the hearbeat network to like a local hub for the nodes or
to the same public network ( dont know if im stating this right) ? Whats
best practise to set up the private hearbeat ?But networks, I am assuming you mean network cards on the server. It is
recommend two; one for the Live check and one for public traffic on the node.
Thanks!
--
Mohit K. Gupta
B.Sc. CS, Minor Japanese
MCTS: SQL Server 2005
"Hassan" wrote:
> I know there is a public network for all application communication and one
> for hearbeat .
> Are there any more ?
> Does one connect the hearbeat network to like a local hub for the nodes or
> to the same public network ( dont know if im stating this right) ? Whats
> best practise to set up the private hearbeat ?
>
>
>