• Hi!

    I've just looked at the connect article and MS seem to have closed it as 'by design'.

    My understanding is that a 'seek' always selects one row (either by key or rid).

    A scan may be a scan of any part of an index, which involves more than a 1-row seek.

    In the example given, a composite primary key on two columns is searched using a value for only the first column.

    There is nothing to say that this will always select one row (unless there is another unique constraint or index on the first column alone, I guess). Since the engine is searching a range in the index (all keys with the value given in the first position) it must be reported as a range scan rather than as a seek, even if it only returns on row.

    This makes a kind of weird M$-sense to me.

    Unless I am missing something there?

    Thanks,

    Paul

    GilaMonster (4/2/2009)


    suhasini.m (3/27/2009)


    2. What does Scan Count exactly mean. And howz it related to IO.?

    The scan count is actually not that useful. It should indicate how many times an object is read, however it doesn't do that consistently. Best bet is to completely ignore that and look at the logical IOs

    see - https://connect.microsoft.com/SQLServer/feedback/ViewFeedback.aspx?FeedbackID=387322