P
Paul Koufalis
Guest
From my experience all variable extents are easier to manage and any performance gains related to fixed extents are imaginary EXCEPT in very specific situations or on very old operating systems. For example, imagine your DB grows by 1 Gb per month. That's 30ish Mg a day = 1.x Mg per hour. Or worse, assume your business only runs M-F 9-5: that 1 Gb of growth only occurs during (20 days X 8 hours = 160 hours) which equals about 6.5 Mg per hour (6.5 Mg X 160 hours ~= 1 Gb). On a storage area with 512 blocks per cluster and a 4 Kb blocksize, each extend operation will grow the extent by 2Mg. This means that your system will "suffer" through 3-4 extensions per hour at a cost of a few milliseconds each. I.e. the cost of variable length data extents in this case is less than one second per hour. Variable length AI extents are a different story but I still typically deploy variable length AI extents. The performance hit here is more measurable BUT what you must measure is the real-life effect on your business. In my experience that real-life performance cost is also negligible to the business even if in terms of absolute values it is measurable. Regards,
Continue reading...
Continue reading...