Re: Oracle Vs. SQL Server on NT

On Mon, 23 Mar 1998 19:52:16 GMT, boulkenospam_at_globalnet.co.uk (Keith Boulton) wrote:

>In addition, read commited isolation is implemented by preventing
>reads of uncommited data (according to what I've read) - a page
>cannot be read by another user if changes to it have yet to be
>commited. Likewise, a page that is currently being read cannot be
>updated. Unlike Oracle, readers block writers and vice versa.
>If too many locks are taken out on a table, the lock is escalated to
>table level, so only one user can update the table.

Generally yes. there is also the option of doing a "dirty read" that may or may not display uncommitted data (I would guess not) but can read locked data just fine, at the cost of data reliability. Also, I believe that SQL can curently do row-level read or write locks -- I forget which offhand.

>This is not a big problem if you design the application to cope with
>it by keeping transactions very short and not leaving cursors open
>etc.

Yes. there is also the issue of clustered indexes vs. nonclustered indexes in relation to what kind of operations you expect to be doing most on the table.

-Sean Received on Mon Apr 13 1998 - 00:00:00 CDT