All entries for Tuesday 14 November 2006
November 14, 2006
- interval partition (e.g, interval by month). DBA does not need to manually create partitions any more.( I just spend half day to create new partitions on my several databases last week)
- gather statistics by columns.
Oracle’s Maximum Availability ArchitectureLarry Carpenter, Senior principle product manager of Oracle, talked about Oracle standby databases. He focused on physical standby.
- Tune redo transfer. Redo transfer can be 100 times more efficient by simply spcify several parameters (SDU = 32K, ....)
Effective Oracle monitoring systemJurijs Velikanovs made a very good presentation about monitoring Oracle. He suggests:
- Use Nagios to monitor server availability. My group is actually using it now.
- Use Cacti to monitor Oracle server performance. It can generated a history graph to display various information, e.g, oracle locks, session memory. This is the kind of tools I have been looking for a while.
I take a quick look at cacti this evening and it use Apache/PHP/MySQL (Not good, I am a Apache/Java/Oracle person). We use Ganglia to monitor unix system. Both Cactic and Ganglia use RRDtool for data storage and visualization. Not sure whether I can use Ganglia to achieve same effect as produced by Cactic.
Accountability for system performance
Six sigma : 99.9997%
He also mentioned DBAs are taughted to ignore waits like “SQL*NET message from client” wait. But did not give reason why not. Hope can have a chat with him.
- I have attend conference for four years in a row. Every year I feel my English hearing is improving. Today I understood most part of the presentations.
I also make mental note about the style of the presentation.
- I was later to go back nursery to pick up Selina. Usually they charge £5 for each 10 minutes. I drove carefully but swiftly. Time is money now.