Presentation is loading. Please wait.

Presentation is loading. Please wait.

Anil Desai Austin CodeCamp 2010.

Similar presentations


Presentation on theme: "Anil Desai Austin CodeCamp 2010."— Presentation transcript:

1 Anil Desai Austin CodeCamp 2010

2 Anil Desai Independent consultant (Austin, TX) Author of numerous IT books Instructor, Implementing and Managing SQL Server 2005 (Keystone Learning) Info: or

3 I. Performance Monitoring Overview II. Monitoring Database Servers III. Using SQL Profiler IV. Using the Database Engine Tuning Advisor V. Application Design Tips VI. Managing Processes, Locking, and Deadlocks

4 Developing processes and approaches for performance optimization

5 Best Practices: Optimize for real-world workloads Monitor/review performance regularly Focus on specific issues Establish a baseline Identify bottlenecks Make one change at a time Measure performance Repeat (if desired)

6 System/OS Windows Performance Monitor Alerts (Performance- Based) SQL Server Activity MonitorSQL Profiler / SQL Trace Database Engine Tuning Advisor Dynamic Management Views (DMVs) Query- Level Database Engine Tuning Advisor Query Execution Plans

7 Server-Level Issues Users are reporting database timeouts Intermittent transaction timeouts The server/application seem sluggish Application- Specific Issues Application A is running more slowly than usual The End-of-Month report is taking too long to run. Other Issues Ad-hoc reports are running slowly CPU, memory, disk, or network alerts are being generated

8 Using SQL Server tools and features to monitor database activity

9 Available in all current versions of Windows Statistics are organized into: Objects Counters Instances Data Collector Sets Windows Vista / Windows 7 / Windows Server 2008 Used to report on performance data that is collected over time Includes built-in System Diagnostics and System Performance collectors and reports

10 Backup Device Device throughput Bytes/sec Buffer Manager Buffer cache hit ratio Page reads / sec Buffer Partition Free Pages Cache Manager Cache Hit Ratio Databases Active Transactions Data File Size Log Growths Percent Log Used Transactions / sec Anil Desai10 General Statistics Logins | logouts / sec User Connections Latches Latch waits / sec Locks Lock requests / sec Lock waits / sec Memory Manager Everything! Replication Depends on configuration SQL Statistics Batch requests / sec SQL compilations / sec

11 SQL Server Activity Monitor Provides a quick overview of database server activity CPU, Processes, Resource Waits and Disk I/O Recent Expensive Queries SQL Server Management Studio Reports Quick overview of SQL Server usage Can export to Excel or PDF

12 Server-Level ReportsDatabase-Level Reports Server Dashboard Memory Consumption Activity – All Block Transactions Activity – Top Sessions Performance – Batch Execution Statistics Performance – Top Queries by Average CPU Object Execution Statistics Disk Usage All Transactions All Blocking Transactions Index Usage Statistics Top Transactions by Age Schema Changes History

13 Windows Event Logs / Event Viewer Application and System Event Logs SQL Server Management Studio SQL Server Logs Can configure max. # of log files SQL Server Agent Error logs Can configure logging levels (Errors, Warnings, Information) Using the Log File Viewer Can Export / Load log information Can search for specific errors/messages

14 Purpose: Monitoring and troubleshooting View server state and performance details Returns relational result sets (use standard SELECT statements) Full list can be viewed in Views System Views section of the properties of the database Scopes: Server level Database level

15 sys.dm_exec_requests Query Execution / Processes sys.DM_DB_File_Space_Usage Storage Engine sys.DM_DB_Index_Operational_Stats sys.DM_DB_Index_Physical_Stats Indexes sys.DM_IO_Pending_IO_Requests sys.DM_IO_Virtual_File_Stats Disk I/O Schema information | Common Language Runtime (CLR) Database Mirror / Clustering | Replication Other Categories

16 Data Collected: System Information (MSINFO) Windows Event Logs SQL Server configuration Command-Line Utility (SQLDiag.exe) Stores output to files Configuration file: SQLDiag.xml Can run as a service (/R) Can run in continuous mode

17 Monitoring SQL Server Activity

18 Purpose / Features: GUI for managing SQL Trace Monitor important events Capture performance data / resource usage Replaying of workloads / transactions Identifying performance bottlenecks Correlation of data with System Monitor Workloads for Database Tuning Advisor Examples: Generate a list of the 100 slowest queries Monitor all failed logins (Security)

19 SQL Profiler Terminology Trace Definitions Events Columns Filters Creating and Managing SQL Traces SQL Profiler (GUI) System Stored Procedures (Transact-SQL) Trace Templates (Built-In) Standard (Default), SP_Counts TSQL, TSQL_Duration, TSQL_Grouped,TSQL_Replay, TSQL_SPs Tuning

20 Anil Desai 20 Trace A set of events, data columns and filters that specify what data should be collected Data can be saved to a file or a database table Trace File Trace data that is saved to a binary file Default extension is.trc Trace Table A SQL Server database table in which trace information is stored Profiler will automatically create the structure of this table when you start running a new trace Trace Template Saved specifications that can be used as the basis for new traces E.g., an environment may have a Security Monitoring template, a CRM Application Performance, etc. Default extension is.tdf

21 Groupings: Event Categories Event Classes Events Examples: TSQL Stored Procedures Performance Errors and Warnings Security auditing Event Categories Event Classes Events

22 Specifies the details to be monitored/recorded Configuring columns Columns can be ordered and grouped Values can be filtered Examples of Columns: StartTime / EndTime TextData Duration Resource Usage (CPU, Reads, Writes) Information: User, Database, App. Names

23 Interactive Good for live monitoring of small sets of data Trace Files (*.trc) Can enable file rollover based on size Server processes trace data option Trace table Will automatically create the table Can set maximum number of rows Scheduling of traces (stop time)

24 Launching SQL Profiler Connecting to a database instance Configuring output options Create a trace definition Specifying events, columns, and filters Running and viewing a trace

25 Creating new templates using SQL Profiler Scripting trace definitions sp_trace_create sp_trace_setfilter sp_trace_GenerateEvent sp_trace_SetEvent sp_trace_SetStatus Extracting SQL Server Events Transact-SQL Events ShowPlan Events Deadlock Events

26 Purpose / Goal: Correlate server performance with database performance Process: Define and start a counter log Define and start a SQL Profiler trace Import Performance Data in SQL Profiler Required Trace properties StartTime EndTime

27 Analyzing workloads to optimize physical database structures

28 Reviews sample workloads and makes performance recommendations Evaluates Physical Design Structures (PDS) Indexes (clustered, non-clustered) Indexed Views Partitions Numerous analysis options Output Generates modification scripts Generates Reports for later analysis

29 Files Transact-SQL Files XML Files Should represent commonly-used queries SQL Profiler Trace Files / Tables Use Tuning built-in trace template Events: Transact-SQL Batch Remote Procedure Call (RPC) Columns: Event Class and Text Data

30 Limit tuning time Tuning Options Allowed Physical Design Structures (PDS) Keep all/specific existing objects Maximum storage space Online or offline recommendations Partitioning

31 Reports can be exported to XML files Report Examples: Workload analysis Column accessTable access View-Table Relations Statement costEvent frequency Index Usage (current / recommended)

32 Process: Generate a workload (file or table) Select tuning options Run the analysis View reports Save and/or apply recommendations Running the DTA: Database Engine Tuning Advisor Application (GUI) Dta.exe command-line utility

33 Practical ways to improve database application performance

34 Create an abstraction layer between the database and the presentation code Separates presentation and logic (esp. in Web Apps) Example: ADO.NET Datasets Database design: Understand typical use-cases before designing the database Create and enforce naming conventions Balance write (OLTP) and read (reporting) performance requirements Use strategic denormalization Anil Desai 34

35 Never include actions that require user input within a transaction Use connection pooling, whenever possible Open connections late and close them early Avoid unnecessary server round-trips Use client-side caching whenever possible Optimistic concurrency Pessimistic concurrency Distribute processing Some operations are more efficient on the DB server (e.g., aggregations, sorting, etc.)

36 Troubleshooting common SQL Server performance problems

37 Processes Interactive users SQL Server Management Studio Applications (Connection Pooling) SQL Profiler Database Engine Tuning Advisor Replication Service Broker Process IDs < 50 are system-related

38 SQL Server Activity Monitor Processes (connected users) Locks (by Process / by Object) Filtering options Auto-refresh option System Stored Procedures / Views Sys.DM_Exec_Sessions Sys.DM_Exec_Requests Sys.SysProcesses sp_who / sp_who2

39 Process Information Current Process ID: Session Options: DBCC USEROPTIONS Killing Processes KILL ProcessID [WITH STATUSONLY] Viewing Last Activity DBCC INPUTBUFFER(ProcessID) DBCC OUTPUTBUFFER(ProcessID)

40 Coordinates multiple accesses to the same data Ensures ACID Properties for transactions (Atomic, Consistent, Independent, Durable) Contention can reduce performance Locking granularity: Row-Level, Page-Level, Table-Level, etc. Lock Modes: Shared, Exclusive, etc. Lock escalation

41 Blocking When transaction(s) must wait for a lock on a resource LOCK_TIMEOUT setting (default = wait forever) Locking Models: Pessimistic Optimistic

42 Balance of concurrency (performance) vs. consistency Affects SELECT queries SET TRANSACTION ISOLATION LEVEL Transaction Isolation Levels READ UNCOMMITTED READ COMMITTED (default) REPEATABLE READ SERIALIZABLE SNAPSHOT Row-Versioning: ALLOW_SNAPSHOT_ISOLATION READ_COMMITTED_SNAPSHOT

43 Activity Monitor SQL Profiler Locks Event Category System Monitor: SQL Server Locks Object System Views Sys.DM_Tran_Locks Sys.DM_Exec_Requests System Stored Procedures sp_Lock

44 Deadlocks: Two or more tasks permanently block each other based on resource locks Default resolution is within 5 seconds Deadlock victim Transaction is rolled-back Process receives a 1205 error Example: Process 1 locks the Customers table and requires access to the Orders Table Process 2 locks the Orders table and requires access to the Customers Table

45 Minimize transaction times Commit / Rollback transactions as quickly as possibly Avoid user-related time within a transaction Access objects in a consistent order Change the transaction isolation level Use a lower level isolation level, if appropriate Use snapshot-based isolation levels

46 Deadlock priorities: SET DEADLOCK_PRIORITY (LOW, NORMAL, HIGH, integer) Deadlock resolution: Lower priority is killed first If equal priorities, least expensive transaction becomes the victim Application or user should attempt to re-run the transaction

47 SQL Server Error Log SQL Profiler Locks Event Category Lock:Deadlock Chain Lock:Deadlock Deadlock Graph Events Extraction Trace Property Export deadlock XML (.xdl) file Viewing Deadlock Files SQL Server Management Studio (File Open SQL Deadlock Files (*.xdl)

48

49 For more information:


Download ppt "Anil Desai Austin CodeCamp 2010."

Similar presentations


Ads by Google