| Author |
Message |
ebferro
Guest
|
 Posted:
Fri Dec 10, 2004 3:05 am Post subject:
Cluster Capacity |
|
|
This is a question for all of the cluster experts out there. We recently
acquired 3 Dell Poweredge 2650's with Dual 3.06 gig 1MB Xeon processors and 2
gigs of ram. Two of these machines, I have clustered together and installed
a SQL Server database on the cluster. We did this primarily for higher
availability, not because our old database server (a 2.4 gig Dual Xeon with 2
gigs of ram machine) was being challenged. We also have an old 500Mhz Dual
Xeon that we're using as a file server. Again, I don't see any bottlenecks
on this machine, either. My question to the experts is this. Given that
neither of the old machines is being taxed, do you think that if we put both
file serving and database on the cluster we'd notice a degradation in
performance? Does anyone know how we might check this prior to doing it?
In addition to that question, I've got another. If I were to take the third
machine and add it to my cluster so I now have three machines instead of two,
would the workload be spread over two machines (N+1) rather than all of the
burden being placed on one machine in a two node cluster? I guess this
question boils down to if I add a third node, do I get an
active/active/passive cluster rather than an active/passive cluster using
only two nodes.
Thanks in advance for any advice that can be offered. |
|
| Back to top |
|
 |
Russ Kaufmann [MCT]
Guest
|
| Posted:
Sat Dec 11, 2004 2:17 am Post subject:
Re: Cluster Capacity |
|
|
"ebferro" <ebferro@discussions.microsoft.com> wrote in message
news:C1ABE07C-210F-49B3-88A5-09486E3EA865@microsoft.com...
| Quote: | This is a question for all of the cluster experts out there. We recently
acquired 3 Dell Poweredge 2650's with Dual 3.06 gig 1MB Xeon processors
and 2
gigs of ram. Two of these machines, I have clustered together and
installed
a SQL Server database on the cluster. We did this primarily for higher
availability, not because our old database server (a 2.4 gig Dual Xeon
with 2
gigs of ram machine) was being challenged. We also have an old 500Mhz
Dual
Xeon that we're using as a file server. Again, I don't see any
bottlenecks
on this machine, either. My question to the experts is this. Given that
neither of the old machines is being taxed, do you think that if we put
both
file serving and database on the cluster we'd notice a degradation in
performance?
|
Maybe, this is why we test. I am guessing it will not be a problem, but
can't say for sure without monitoring performance on both for a considerable
time.
| Quote: | Does anyone know how we might check this prior to doing it?
|
Look at resource utilization on both servers.
| Quote: | In addition to that question, I've got another. If I were to take the
third
machine and add it to my cluster so I now have three machines instead of
two,
would the workload be spread over two machines (N+1) rather than all of
the
burden being placed on one machine in a two node cluster?
|
It could be. It depends on whether you create multiple instances of SQL, for
example.
You could attack this two ways (OK, more, but here are two anyway)
1. Use a 2 node cluster, put SQL on node1 and put File/Print on node2.
2. Use a 3 node cluster, put SQL on node1, File and Print on node2, and
configure both to fail to node3. |
|
| Back to top |
|
|
ebferro
Guest
|
| Posted:
Sat Dec 11, 2004 2:59 am Post subject:
Re: Cluster Capacity |
|
|
Russ:
Thanks for the feedback. Excuse me if these are dumb questions but I'm a
cluster newbie so I've got a way to go. Some light bulbs may be starting to
go off now, though so I may be making progress.
When you say 'Put SQL on node 1 and File/Print on node 2', does that mean
that I should install SQL on node 1 only and install File/Print on node 2?
Or, do you mean that SQL should be installed to the cluster while on node 1
so that node 1 'owns' it so to speak and that File/Print services should be
installed to the cluster from node 2 so that node 2 'owns' those services?
That way, each server does some work but if there's a failure the services
move to whichever node is still running. I guess what's confusing to me is
that we have this cluster. I figured when we installed something to the
cluster, it became node independent meaning that the cluster ran it, not the
computer from which or to which it was installed. Am I looking at this the
wrong way, Russ?
TIA.
Ernie
"Russ Kaufmann [MCT]" wrote:
| Quote: | "ebferro" <ebferro@discussions.microsoft.com> wrote in message
news:C1ABE07C-210F-49B3-88A5-09486E3EA865@microsoft.com...
This is a question for all of the cluster experts out there. We recently
acquired 3 Dell Poweredge 2650's with Dual 3.06 gig 1MB Xeon processors
and 2
gigs of ram. Two of these machines, I have clustered together and
installed
a SQL Server database on the cluster. We did this primarily for higher
availability, not because our old database server (a 2.4 gig Dual Xeon
with 2
gigs of ram machine) was being challenged. We also have an old 500Mhz
Dual
Xeon that we're using as a file server. Again, I don't see any
bottlenecks
on this machine, either. My question to the experts is this. Given that
neither of the old machines is being taxed, do you think that if we put
both
file serving and database on the cluster we'd notice a degradation in
performance?
Maybe, this is why we test. I am guessing it will not be a problem, but
can't say for sure without monitoring performance on both for a considerable
time.
Does anyone know how we might check this prior to doing it?
Look at resource utilization on both servers.
In addition to that question, I've got another. If I were to take the
third
machine and add it to my cluster so I now have three machines instead of
two,
would the workload be spread over two machines (N+1) rather than all of
the
burden being placed on one machine in a two node cluster?
It could be. It depends on whether you create multiple instances of SQL, for
example.
You could attack this two ways (OK, more, but here are two anyway)
1. Use a 2 node cluster, put SQL on node1 and put File/Print on node2.
2. Use a 3 node cluster, put SQL on node1, File and Print on node2, and
configure both to fail to node3.
|
|
|
| Back to top |
|
|
|
|
|
|
FAQ
Memberlist
Register
Profile
Log in to check your private messages
Log in