distributed/linked queries have a high obverhead, if at all possible I only make rpc calls - I would have a proc ( query ) on the remote server and use...

yup - check out BOL for a starting implementation of how to do selective re-indexing - as you say it's a matter of availble time and complexity of putting a...

Partitioned views will solve your problem ( assuming you get the partitions right ). Note that SQL2005 supports transparent table partitions.  57 million rows / 30 gb isn't too big...

You may find increasing memory will drop cpu .. things I've done which have reduced cpu load ( depends upon your apps of course ):-

Added memory

Added secondary indexes to stop...

I'm not sure as to exactly what concerns you, you will often see lots of pagelatch_ex, as these are required for data changes. It's only if you have contention/blocking that...

It's only if you modify the primary key you run into problems. secondary indexes are unaffected.

If you can figure out the T-SQL ( rather than the gui ) making changes...

I would personally never use a maint plan - always preferring to write the jobs in clear tsql - beware that sysmaint.exe works as an external process to sql server....

It's a long time since I used 7.0 but I seem to remeber it works the same as 2k and will restore the database with the original sort order (...

It's fairly easy - but care is required - there are a number of ms articles

Microsoft Knowledge Base Article - 263712

don't have the exact link to hand as I...

Yup .. but creates a cursor underneath the covers. I did the proc(s) so I could bind it into maint jobs ( without the maint plans ) and you know...

Yes you have to reformat to change block and stripe size - that's why I never had time to really test it out.

I have different views on SANs which I...

Not quite sure what your question is - distributed transactions have an overhead but rpc calls are usually pretty quick. Any number of issues can cause different run times ona...

I continue to run 1204, 1205 and 3605 trace flags on my prod cluster ( did have lots of deadlock issues ) and this was the absolute way to get...

It's claimed a larger block size will give improved performance, I've never been able to prove it ( takes too long to set up and is difficult to benchmark )...

I assume you mean one command vs multiple commands ?  I'd really suggest you follow the route shown in BOL for showcontig which shows how to do selective rebuilds.

I use...