Hardware requirements for SQL Server 7.0 (Transaction Analysis) 1

[bnath001]

Hello,

I am trying to buy a desktop computer for doing Transactional Analysis assignments. I wanted to find out what kind of machine should I be looking for.

I want to use SQL Server 7.0 and the data (for analysis) could be around 5 to 10 GB in size approximately. I may not use this as Web server or fun running Business critical applications. this would be used for Data Analyis (running lot of queries in SQL server with Windows 2000)

Can someone suggest the hardware configuration I should be looking for?

Thank you
nath

 

[foxdev]

A lot of things will work, but obviously more $$ = more performance.

If you can afford it, get a dual-processor (or more) box with a RAID system and 512+ Mb RAM. The multiple processors work especially well if the bulk of the load is for queries, rather than inserting and updating transactions. The RAID drives will allow disk reads to be made in parallel. And the extra RAM allows the server to keep more of the data in memory, and would allow you to "pin" more tables in memory.

Robert Bradley

http://www.foxdev.com/

 

[TomSark]

I would not pin tables in memory, ever. Microsoft frowns upon this and so do I. You should allow the buffer management of SQL Server to do it's job.. it does it very well.

You will get the best performance by having lots of memory, multiple CPUs, fast disk I/O, with multiple controllers, and proper database architecture. Also, make sure your hardware is on the Windows 2000 HCL (and NT 4 HCL if applicable).

Tom

 

[foxdev]

While I agree that table pinning should rarely be done, and only done when a careful analysis shows it to be advisable, I disagree with your blanket statement that Microsoft frowns on it. Your point that the SQL Server cache manager does a dang good job of managing memory is well-taken, but in the BOL Microsoft says:

DBCC PINTABLE is best used to keep small, frequently referenced tables in memory. [...] Although DBCC PINTABLE can provide performance improvements, it must be used with care.

Well, OK: pretty cautionary words...but I'd say its more of a furrowed brow than a frown ;-)

What I had in mind when I wrote that were the small, relatively static lookup tables that might be joined in queries for readability, not larger transactional tables.

Robert Bradley

http://www.foxdev.com/

 

[TomSark]

And by the way, proper index architecture and good written queries are the probably the most important thing you can do to having speedy, healthy applications (still waiting to find out how I can put a smiley face in my messages.. hint hint).

Tom

 

[foxdev]

still waiting to find out how I can put a smiley face in my messages

Click on the emoticons/smileys link at the bottom of the page.


Robert Bradley

http://www.foxdev.com/

 

[TomSark]

Thanks ;-)

Tom