June 6, 2012 at 1:36 pm
We currently have 4 SQL Server Instances on a Stand-alone machine running Windows Server 2008 R2 Enterprise and SQL Server 2005 SP3 Standard Edition and the following hardware configuration:
2 processors - Intel Xeon CPU x56660 @ 2.80GHz (Hexacore)
32GB RAM
64bit
300GB of local Storage
350GB of SAN Storage
One of the SQL Server Instances is a Sharepoint Instance and in order to take advantage of some of the enhancements in moving the Sharepoint Backend to SQL Server 2008 R2 Enterprise Edition (enhanced BLOB storage in particular) I've been asked to make a case for a 2 Node HA Cluster and SQL Server 2008 R2 Enterprise Edition and put a proposal together.
making a case for SQL Server 2008 R2 Enterprise is not really my problem. The problem lies in trying to find good information on the actual hardware requirements for the 2 Node cluster with 4 SQL Server Instances - Processors, Memory, etc.. what we have now is adequate but do we need to ramp up the processors and memory for sql server 2008 R2 enterprise and the clustering? Can anyone help me out? I've googled this till my fingers hurt : )
June 6, 2012 at 2:16 pm
Hi Jpotucek,
The moment you think about an Active-Active Cluster you will need to think about double the resources on each node of the 2 node HA cluster, at any given point.
for instance, you mentioned:-
you are happy with 2 processors - Intel Xeon CPU x56660 @ 2.80GHz (Hexacore) for 4 instances of SQL.
That would mean you will need 2 servers with 2 processors - Intel Xeon CPU x56660 @ 2.80GHz (Hexacore)
for each nodes of the cluster. Even better get 2 dual/quad core processors on each nodes of the cluster. Choose wisely as you will not find a better opportunity than this to upgrade your hardware; at the same time balance the spec. But here is the maths with the present numbers you have.
You currently have 32GB RAM divided between 4 instances on one server. I would estimate you are probably using upto 28GB RAM between 4 instances; as anything allocated to SQL above that would leave very less RAM for the OS and would cause bottleneck.
You will need 32GB RAM on each nodes i.e., a total of 64GB RAM on the cluster. Again as I am not aware of the load and requirements of the server, I can only provide estimates based on the present numbers.
Since this a two node cluster with 4 instances of SQL, at any given time you would only be using half the resources on the server after allocating resources for the OS. The remaining half of the resources will need be to set aside and ready to be used when the failover happens.
Win_node_A 2/4 processors - Intel Xeon CPU x56660 @ 2.80GHz (Hexacore) with 32 GB RAM
SQL_Instance1(Active) -- set the Min mem setting to 2GB and Max mem setting to 7GB
SQL_Instance2(Passive) -- set the Min mem setting to 2GB and Max mem setting to 7GB
SQL_Instance3(Active) -- set the Min mem setting to 2GB and Max mem setting to 7GB
SQL_Instance4(Passive) -- set the Min mem setting to 2GB and Max mem setting to 7GB
i.e., a max of 28GB of RAM is allocated to 4 instances of SQL Servers on Win_Node_A; leaving 4GB memory for the OS on Win_Node_A
Win_node_B 2/4 processors - Intel Xeon CPU x56660 @ 2.80GHz (Hexacore) with 32 GB RAM
SQL_Instance1(Passive) -- set the Min mem setting to 2GB and Max mem setting to 7GB
SQL_Instance2(Active) -- set the Min mem setting to 2GB and Max mem setting to 7GB
SQL_Instance3(Passive) -- set the Min mem setting to 2GB and Max mem setting to 7GB
SQL_Instance4(Active) -- set the Min mem setting to 2GB and Max mem setting to 7GB
i.e., a max of 28GB of RAM is allocated to 4 instances of SQL Servers on Win_Node_B; leaving 4GB memory for the OS on Win_Node_B
=======================================================================================
Visit my technical reference; you might find some of your issues already documented.
June 6, 2012 at 2:42 pm
As already mentioned, take this opportunity to ramp up the CPU and RAM.
Aside from that, would there be any mileage\possibility consolidating the remaining instances?
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
June 7, 2012 at 3:17 am
Thank you for the replies. It goes without saying that i will try and ramp up the CPU and memory on any new solution, I just didn't know what the additional requirements were specific to the windows clustering - sounds like what I have might be sufficient but a 'ramp up' would be a nice thing to have. Have to think about the $$ because anything I do in the cluster is 'times 2' - hardware, software and licensing, fiber cables to the SAN etc...
as far as consolidating the instances.. Sharepoint really should be in it's own Named Instance and I have another application that requires it's own instance because of a special SQL Server collation that it requires. I will look at consolidating the remaining to instances into the default instance.
June 7, 2012 at 4:44 am
Jpotucek (6/7/2012)
Sharepoint really should be in it's own Named Instance and I have another application that requires it's own instance because of a special SQL Server collation that it requires.
That's an entirely valid reason for separate instances
Jpotucek (6/7/2012)
consolidating the remaining to instances into the default instance.
😉
-----------------------------------------------------------------------------------------------------------
"Ya can't make an omelette without breaking just a few eggs" 😉
June 8, 2012 at 7:10 am
Makes perfect sense to me !
=======================================================================================
Visit my technical reference; you might find some of your issues already documented.
Viewing 6 posts - 1 through 5 (of 5 total)
You must be logged in to reply to this topic. Login to reply