Golden Opportunity at work

Soldato
Joined
20 Oct 2002
Posts
3,467
Location
London, UK
Hey guys,

My boss has just told me that our team has been offered 3 old-ish Blade servers for our SQL Server use, which is hugely better than the 32-bit VMWare based SQL Server we are provided with right now.

I can build a home PC, troubleshoot, phase-cool overclock, etc but I respect that enterprise server solutions are a whole different ball game :p

We are being given 2 of these Poweredge systems:
Xeon X5650, 64GB RAM, 120GB HDD

And 1 of these:
Xeon X5650, 32GB RAM, 120GB HDD

What we want to do is perhaps combine these into one "uber" powerful SQL Server instance (maybe only the first two as a production environment and have the 3rd as a pre-prod). We have IT support etc but before I talk to them I was wondering I could get some thoughts from you guys who have experience in this sort of thing.

Our main uses are quite ad-hoc but heavily I/O intensive analytics, with an informal IIS server to boot for web based publishing (small-ish audience, no more than 1000 people total, most of them Exec with little tech knowledge). We'll be processing & building OLAP cubes on it in the near future also.

We will probably keep our existing virtual machine as an experimental sandbox.

Any advice would be hugely appreciated :cool:
 
Last edited:
'Production' infers service availability requirements, so a SQL cluster is probably a sensible thing. Having another lone blade as a stand-alone staging makes sense as well.

I don't have much OLAP experience, but I'd check that the hardware performs how you want it to before investing time and effort.

Only other thing to note is that SQL clusters require a domain controller, which can't be on either cluster node, and SAN storage.
 
I'm not a DBA but I don't believe there are any methods for load balancing SQL servers. A cluster will only provide higher availability by providing a fail over.

Maybe you could replicate the database from 1 server to the second and base all activity that needs read only access (reporting) to read from server 2?
 
Hmm I was afraid that might be the case, I'm wondering then if it's worth just combining all the servers into a VMWare farm and creating virtual machines where applicable...
 
I'm not a DBA but I don't believe there are any methods for load balancing SQL servers. A cluster will only provide higher availability by providing a fail over.

Maybe you could replicate the database from 1 server to the second and base all activity that needs read only access (reporting) to read from server 2?

Load balancing would generally be done in the application rather than at the DB layer.
 
also, it looks like those boxes just have a single disk? if so and you dont have anything better then those uber processors are just going to sit waiting for IO. SQL is hugely IO sensitive and getting a good amount of IO capability is usually a number 1 priority followed by memory. CPU comes a poor 3rd.
 
Back
Top Bottom