Posted by Erland Sommarskog on 12/06/07 22:16

Neil (nospam@nospam.net) writes:
> OK, that's fine. But my POV is that rebooting in general, and resetting
> the system memory, is a good thing to do. I take it that you disagree?

I do. Unless there is some known issues, there is rarely any reason to
reboot SQL Server regularly. OK, so you have SQL 7, and this version is a
bit old. Nevertheless, it has the same basic architecture as subsequent
SQL Server versions. The predecessors 6.x were less stable. (In 6.0 there
was a bug that compelled you to restart the bug about every 48th day.
There was some counter somewhere that wrapped around after 49 days.)

As for rebooting helping performance, it may be the other way round.
When you reboot, the entire plan cache will be thrown away, and until
the plan has built up, the server will spend more time compiling. But
that's the only thing. Say that you have a table with date column,
into which the current date is typically inserted, and queries against
this table are typically for the recent days. If the query is a stored
procedure or is parameterised, the plan will be built using that date.
But if statistics are out of date, the estimates will be wrong, and you
may get a bad plan. Had you never restarted the server, the old good
plan would have remained in cache.

But, OK, the reverse can happen: the old plan was bad, but statistics
have been updated recently, so now you get a good plan.

As for your problem the other day, you did not include any error message,
but I would guess that it was a temporary network problem, and it would
have gone away even if you had not restarted SQL Server.

--
Erland Sommarskog, SQL Server MVP, esquel@sommarskog.se

Books Online for SQL Server 2005 at
http://www.microsoft.com/technet/prodtechnol/sql/2005/downloads/books.mspx
Books Online for SQL Server 2000 at
http://www.microsoft.com/sql/prodinfo/previousversions/books.mspx

[Back to original message]