Reducing CPU problem caused due to high buffer gets (after large delete)

Reducing CPU problem caused due to high buffer gets.
Couple of month’s back I was working on tuning one of the telecom databases which was occupying almost 90% of the CPU in peak times.

While interpreting the statspack report I found that one of their queries is consuming more CPU and it has more buffer gets, their count running into some crores.

I was looking at that particular query and surprised to find that it has only 20 records in that table and performing a full table scan. Based on this I ran the following query

select blocks,num_rows from dba_tables where owner='XYZ';

to get the no of blocks the table has occupied and found that it has occupied 1500 blocks for 20 rows. Each time the query runs, it is scanning 1200 blocks for 20 records which is the cause for the high CPU consumption.

The CPU came back to 20% after recreating the table where it occupied 2 blocks for 20 records.

The reason for this is that, the table used to have millions of records earlier and they have performed a delete operation, which is the cause for blocks not being released after the delete.
 
發佈了19 篇原創文章 · 獲贊 0 · 訪問量 7萬+
發表評論
所有評論
還沒有人評論,想成為第一個評論的人麼? 請在上方評論欄輸入並且點擊發布.
相關文章