oracle internal :VIEW: X$KCBLDRHIST - Direct Read HISTory
| WebIV:View NOTE:159900.1 |   |
 |
 |
 |
Article-ID: <Note:159900.1> Alias: VIEW:X$KCBLDRHIST Circulation: PUBLISHED (INTERNAL) ***Oracle Confidential - Internal Use Only*** Folder: server.Internals.General Topic: ** X$ Table Definitions Title: VIEW: X$KCBLDRHIST - Direct Read HISTory - (9.0) Document-Type: REFERENCE Impact: LOW Skill-Level: NOVICE Server-Version: 09.00 Updated-Date: 05-OCT-2001 05:27:02 References: Shared-Refs: Authors: MEPRESTO.US Attachments: NONE Content-Type: TEXT/PLAIN Products: 5/RDBMS (9.0); View: X$KCBLDRHIST [K]ernel [C]ache [B]uffer [L]oad management [d]irect [r]ead [hist]ory Column Type Description -------- ---- -------- ADDR RAW(4|8) address of this row/entry in the array or SGA INDX NUMBER index number of this row in the fixed table array INST_ID NUMBER oracle instance number DRTHRUPT NUMBER throughput of direct reads TOTLSLTS NUMBER Total number of slots as seen by CKPT INCREASE NUMBER Increase bit KCBLHV_STAY 0 Do not increase number of slots KCBLHV_INCR 1 Increase number of slots KCBLHV_SYNC 2 Syncronize mode REALSLTS NUMBER Total number of slots as seen by the PQ slaves Notes from source: One copy of the history vector is allocated per instance. Each entry corresponds to a different time snapshot. Every 3 seconds, the next entry is updated. Each entry contains the following information: - Throughput of direct reads during the last 3 seconds expressed in blocks per second. - Total number of slots (1). The background process every 3 seconds adds up the number of slots in all active sessions. This is possible since every session maintains statistics about its number of slots. - Total number of slots (2). Each slave running adaptive direct reads adds regularly its number of slots to the history vector. The reason for having two different variables representing the same logical entity is because either of those two variables alone is not accurate enough. For example, if just before the end of the 3 second interval, a new slave started then (1) will not be accurate. On the other hand, if a slave just missed the chance to add its number of slots in the last entry, then (2) is wrong. So before any decision is taken, we make sure that (1) is identical to (2). - Increase bit. This bit indicates the decision taken by the background process for the next 3 seconds interval based on the previous history entries. Check the note in kcblclc for more information about the increase bit. KCBLHSZ 1000 Size of History Vector in number of rows KCBLHSLTRT 25 Minimum percentage increase in number of slots Minimum percentage increase in throughput expressed in percentage increase in the number of slots. For example, if we have a 40% increase in the number of slots, then the minimum increase in throughput needs to be 40% * 25% = 10%. So this number correlates the increase in number of slots with the throughput. The minimum increase in throughput is used by the background process as will be seen later. The latter compares the current increase in throughput with the minimum increase to determine whether to increase the number of slots in the system or not. KCBLHTHPTRT 25 Maximum IO Wait to Cumulative time ratio where slave is CPU bound expressed in percentage. If that IO wait to cumulative time ratio is greater than this number, then the slave is considered IO bound and would increase number of slots if background process indicates to do so. Otherwise, the slave is considered CPU bound and consequently ignores all decisions taken by the background process. KCBLHCPUBDRT 10 NOTE ON PERSISTENT AND NONPERSISTENT STATE FOR ADAPTIVE READS: The state for adaptive direct reads is composed of a persistent and nonpersistent state. Persistent state means that kcbl state is maintained across uses of a load descriptor. In other words, the state is maintained across kcbldrini calls. The reason for that is that sometimes kcbldrini is called multiple times for the same table or index... We don't want to restart the adaptation from scratch everytime kcbldrini is called. Instead we would like to maintain some state so that in case kcbldrini is called again on that same object, no redundant adaptation is needed.
