G
George Potemkin
Guest
> 75%+ of all CREATE activity occurs in just 3 tables on our 9.1 DB (i.e. Type 1 extents) Do you have the statistics for these tables?: _TableStat-update vs _TableStat-create vs _TableStat-delete > Would it be worth having these in their own extent to reduce Scatter? My thinking here is that all this delete activity is peppering the DB with gaps like Swiss cheese leading to high scatter over time. How long are RM chains of these tables? It's more important than their scatter factors. > There is a KB which also says to subtract 100 from the DB block size to accommodate the block header information 100 bytes is close to the size of RM block headers in storage area type 2. If one cares about such small pieces of block space then the toss limit (300 bytes for 8K) should be taken into account as well. But I wonder why we need to calculate the best number of records per block with a lot of the decimals and then round the value up to a power of two (not even to a nearest integer)?
Continue reading...
Continue reading...