Load 1TB in less than 1 hour

http://blogs.msdn.com/sqlcat/archive/2006/05/19/602142.aspx

Excerpt from the article-

BEST PRACTICES and LESSIONS LEARNED

Here is a list of things we learned during this project that should be helpful in your projects.

• Run as many load processes as you have available CPUs. If you have 32 CPUs, run 32 parallel loads. If you have 8 CPUs, run 8 parallel loads.
• If you have control over the creation of your input files, make them of a size that is evenly divisible by the number of load threads you want to run in parallel. Also make sure all records belong to one partition if you want to use the switch partition strategy.
• Use BULK insert instead of BCP if you are running the process on the SQL Server machine.
• Use table partitioning to gain another 8-10%, but only if your input files are GUARANTEED to match your partitioning function, meaning that all records in one file must be in the same partition.
• Use TABLOCK to avoid row at a time locking.
• Use ROWS PER BATCH = 2500, or something near this if you are importing multiple streams into one table.

Gaurav
The views expressed here are those of the author and no one else. There are no warranties as to the reliability or accuracy of anything presented here.
http://blogs.msdn.com/gauravbi/default.aspx

this is a good tryout. do you think it can be done on SQL server 2000?

Its more like a blog entry.

This is blog entry by SQL Server Customer Advisory Team which is one of the main interface to external customers of SQL Server.

I would think that the content is as credible as any other white paper from Microsoft.

Gaurav
The views expressed here are those of the author and no one else. There are no warranties as to the reliability or accuracy of anything presented here.
http://blogs.msdn.com/gauravbi/default.aspx

>> Run as many load processes as you have available CPUs. If you have 32 CPUs, run 32 parallel loads. If you have 8 CPUs, run 8 parallel loads.

Won't this be limited by disk performance rather than cpu?

1TB in 50 min works out to 330MB/sec which a little higher than i have seen, but this is much bigger hardware,
any ways, this disk load is would only slightly dent a serious disk system

my suspicion is that it is not a simple matter formatting input data into the SQL Server table format,
i seem to recall that Oracle lets you format data into the native table format for faster loads? or something like this.
so the question is: does any one see the need to load data faster than 330MB/sec?
i have seen the 64P Itanium system do a table scan at 12GB/sec