10G - New Manageability Features

1
10G - New Manageability Features
Presented by Lenka Vanek lenka.vanek_at_quest.com
2
Oracle 10g Manageability

• Active Session History (ASH)
• Contains recent session activity
• Automatic Workload Repository (AWR)
• Infrastructure -gt Central element provides
  services
• Automatic Database Diagnostic Monitoring (ADDM)
• Generate advice based on AWR data
• Server Generated Alerts
• - metrics computation and and threshold
  validation done by Oracle database 10G directly

3
ASH v session History

• Contains recent session activity
• History of vsession_wait records what is
  session waiting for
• Every Second
• Inactive sessions not sampled.
• Design rolling buffer in memory
• Size - between 1M 128M (avg. sample record 600
  bytes)
• --- Algorithm used to estimate ASH buffers
  sizememory_quota max(2 of sga_target, 5 of
  shared_pool_size) / sga_target 0 when AUTO
  SGA is OFF /cpu_quota    2MB ( of
  CPUs)ash_size     min( cpu_quota,
  memory_quota )ash_size     max( 1MB,
  ash_size) / atleast 1MB /ash_size     min(
  128MB, ash_size) / atmost 128MB /

4
ASH

• Provider for ADDM
• vsession and vsession_wait join eliminated
• Prior 10G - sessions experiencing waits were
  generally located by joining the vsession_wait
  view with the vsession view.
• 10G - offers query simplification. All the wait
  event columns from vsession_wait have been added
  to vsession.
• xash
• Vactive_session_history - contains one row for
  each active session per sample
• DBA_HIST_ACTIVE_SESS_HISTORY - contains
  historical data
• the greater the system activity, the smaller the
  number of seconds of session activity that can be
  stored in the circular buffer
• Flushed every 30 minutes or when buffer is full
• MMON every 30 minutes and by MMNL (Memory Monitor
  Light) whenever the buffer is full
• wrhactive_session_history

5
ASH
Statistics ASH vsession SGA
Rolling Buffer Recent History 30 min is just goal
VACTIVE_SESSION_HISTORY
MMON MMNL
AWR Snapshots
6
ASH - Limitations

• Query of vactive_session_history needs a session
• Query of vactive_session_history requires all
  relevant latches in SQL layer
• If system is crippled ASH will impose more
  overhead on these latches

7
AWR Automatic Workload Repository

• AWR is Infrastructure
• Collects, maintains and utilizes statistics
• Two major parts
• In-memory statistics fixed views- V
• WR Schema, Snapshots persistent portion for
  historical analysis.
• SYSAUX tablespace - occupies 63.7 of space
• Process MMON memory monitor disk transfer,
  snapshots purging, retention period

8
AWR
SQLPlus
BG
..
SGA In memory statistics Collections Time model,
wait classes, OS stats, Metrics, SQL
Stats Object Stats
ASH
BG
AWR Snapshots SYSAUX 7 days - Default
V
DBA_
MMON
FG
FG
vsysstat vsql vsegment_statistics vsys_time_mo
del vosstat vevent_name
ADDM
9
AWR STATSPACK

• Foundation for all of the other self-tuning
  features.
• Runs every 30 min
• Provides data for
• ADDM
• Alerts
• Advisors
• Cost Based Optimizer
• End-to-end tracing
• Automatically installed, populated, purged for
  10G only
• Default retention 7 days. This can be changed.

10
AWR and Snapshots

• Stores information in form of Snapshots (similar
  to statspack snapshots, but more precise)
• Snapshot set of data captured at a certain time
• Each time a snapshot is taken, the ADDM is
  triggered to do an analysis of the period
  corresponding to the last two snapshots

Snapshot can be taken manually

• BEGINDBMS_WORKLOAD_REPOSITORY.CREATE_SNAPSHOT
  ()END
• /

11
AWR Base Statistics and Metrics

• Base statistics raw data
• Metrics secondary, derived from base statistics
• Updated by MMON
• Example avg number of physical reads per sec in
  the system in the last 30 minutes
• Tract the rates of change
• Indicators of DB performance
• Deltas of Stats and Events over 15 and 60 seconds
• Max, min, avg., standard deviation over 30 min
• 10 minutes for File IO
• 30 minutes for SQL metric values are constantly
  increasing in x until 30 minute snapshot when it
  is externalized into dba_hist_sqlstat

12
AWR - Metrics

• New Views
• vsysmetric .not a 1 to 1 map of vsysstat,
  there are some new values
• vsessmetric
• vmetricname
• vfilemetric, vwaitclassmetric, veventmetric,
  
• 10G Supports metrics for
• System
• Session
• File
• Wait-event statistics
• The wait event model - steadily gaining ground as
  a good tuning tool
• At any given moment an Oracle process is either
  busy servicing a request or waiting for something
  to happen

13
Wait Event Enhancements

• Formed new wait events classes before too many
  individual events
• Changes to vevent_name - CLASS and CLASS
  columns are added. These columns help to group
  related events while analyzing the wait issues.
• Example - list the events related to IO,
• Understanding the overall health of the database.
• New columns in the vsession and vsession_wait
  views that track the resources sessions are
  waiting for.
• Histograms of wait durations
• Assist in determining whether a wait event is a
  frequent problem that needs addressing or a
  unique event.

14
New Views

• vsystem_wait_class the instance-wide time
  totals for the number of waits and the time spent
  in each class of wait events. Understanding the
  overall health of the database
• vsession_wait_class - the number of waits and
  the time spent in each class of wait event on a
  per session basis.
• vevent_histogram a histogram of the number of
  waits, the maximum wait, and total wait time on a
  per-child cursor basis.
• vfile_histogram a histogram of all single
  block reads on a per-file basis. Determine if the
  bottleneck is a regular or a unique problem.
• vtemp_histogram a histogram of all single
  block reads on a per-tempfile basis.
• vsession_wait_history This view displays the
  last 10 wait events for each active session.

15
Wait Classes Overview

• Administrative (39)
• switch logfile
• rebuild index
• Application (11)
• enqueues
• sqlnet break/reset
• Cluster (113)
• Commit (1)
• Log file Sync
• Concurrency (12)
• Latches cbc, lbc,
• Lib cache locks
• Buffer busy wait

• Configuration (20)
• log file size
• Enqueues ST, HW, ITL
• Latch redo copy,shared pool
• Idle (56)
• Network (25)
• System I/O (19)
• Scheduler (6)
• User I/O (12)
• Other (485)

16
AWR and STATISTICS

• Oracle10g to be either robust or simple
• You can control the set of statistics to capture
  by using STATISTICS_LEVEL parameter.
• If STATISTICS_LEVEL is set to
• BASIC The computation of AWR statistics and all
  self-tuning capabilities are turned off.
• TYPICAL Only part of the statistics are
  collected. They represent what is typically
  needed to monitor the Oracle server behavior.
  DEFAULT.
• ALL All possible statistics are captured.

17
AWR Limitations

• Only for 10G
• Each DB has its own AWR repository
• Cannot provide cross instance analysis
• Overhead on production box
• PL/SQL interface cannot communicate if box is
  down
• STATSPACK cannot be migrated into 10G
• User cannot modify AWR Schema

18
Automatic Database Diagnostic Monitor

• Performance Diagnostics within DB
• Not a monitor perform analysis an Advisor
• Analyzes the AWR data, much the same as a human
  DBA would analyze a STATSPACK report
• Utilizes AWR and publishes report - every hour
• Identify problems - ADDM searches for
  lock-and-latch contention, file I/O bottlenecks
  and SGA shortages, etc
• Proposes solutions relies on Advisor for
  solution

19
ADDM

• Where is the time spent?
• What did DB do?
• Elimination method where is NOT your problem
  TREE Structure
• Uses a tree structure to represent all possible
  tuning issues
• The tree is based on the new wait and time model
  statistics
• Root of this tree represents the symptoms, and
  going down to the leaves
• If time-based threshold is not exceeded for a
  particular node, ADDM prunes the corresponding
  sub-tree

20
Limitation of ADDM

• Same limitation as AWR only for 10G
• Works of a set of rules and these are not
  external of DB. Based out of fixed thresholds.
• Depends on ASH for analysis

21
Advisory Framework

• Advisors are server components - provide useful
  feedback about resource utilization and
  performance
• Automatic Database Diagnostic Monitor (ADDM)
  An advisor for the database instance
• ADDM can call
• SQL Tuning Advisor - tuning advice for a SQL
  statement
• SQL Access Advisor - determines optimal ways to
  access data
• Space Advisor
• Segment Advisor Responsible for space issues
  regarding a database object. It analyzes the
  growth trends.
• Undo Advisor Suggests parameter values and the
  amount of additional space that is needed to
  support flashback for a specified time

22
Advisory Framework

• Memory Advisor MMAN
• PGA Advisor recommends optimal usage of PGA
  memory based on your workload.
• SGA Advisor tuning and recommending SGA size
  depending on pattern of access for the various
  components within the SGA
• Buffer Cache Advisor Predicts cache hit rates
  for buffer access for different sizes of the
  buffer cache.
• Library Cache Advisor Predicts the cursor cache
  hit rate for the library cache for different
  sizes.
• AWR Common Data Source

23
10G Server Generated Alerts

• Threshold and non-threshold alerts
• Tablespace full
• Snap-shot too old, Recovery Area Low On Free
  Space, Resumable Session Suspended
• Notification by page/email/PDA
• Push not pull for efficiency
• Advanced Queue
• Server pushes alerts
• Predefined persistent queue ALERT_QUE owned by
  SYS
• Alerts persist to AWR, review historically
• purged according to Workload Repository snapshot
  purging policy

24
10G Server Generated Alerts

• Most Oracle server-generated alerts are
  configured by setting two threshold values on
  database metrics
• Warning threshold 85
• Critical threshold 97
• Only space-related alerts have thresholds defined
  by default Tablespace Space usage
• There are 161 metrics for which you can define
  thresholds

25
Server Alerts Limitations

• Concept great implementation has a lot to be
  desired
• Only 4 alerts out of the box
• Rest needs to be set up by user
• Avoid false peaks need to set of occurrences.
  This is very difficult in production- How long
  should I wait?

26
QUESTIONS ANSWERS