azure-sql-optimization

Azure SQL Database Optimization

Platform-specific optimization for Azure SQL Database.

Quick Reference

Service Tier Comparison

Tier	Best For	Max Size	Key Features
Basic	Dev/test, light workloads	2 GB	Low cost
Standard	General workloads	1 TB	S0-S12 DTUs
Premium	High I/O, low latency	4 TB	P1-P15 DTUs
General Purpose (vCore)	Most workloads	16 TB	Serverless option
Business Critical	High availability	4 TB	In-memory, read replicas
Hyperscale	Large databases	100 TB	Auto-scaling storage

DTU vs vCore

Aspect	DTU	vCore
Pricing	Bundled resources	Separate compute/storage
Control	Limited	Fine-grained
Reserved capacity	No	Yes (up to 72% savings)
Serverless	No	Yes (General Purpose)
Best for	Simple workloads	Predictable, migrated workloads

Performance Monitoring

Resource Consumption

-- Last 15 minutes (avg 15-second intervals)
SELECT
    end_time,
    avg_cpu_percent,
    avg_data_io_percent,
    avg_log_write_percent,
    avg_memory_usage_percent,
    max_worker_percent,
    max_session_percent
FROM sys.dm_db_resource_stats
ORDER BY end_time DESC;

-- Historical (last 14 days, hourly)
SELECT
    start_time,
    end_time,
    avg_cpu_percent,
    avg_data_io_percent,
    avg_log_write_percent
FROM sys.resource_stats
WHERE database_name = DB_NAME()
ORDER BY start_time DESC;

Query Performance Insight

-- Top CPU consumers last hour
SELECT TOP 20
    qt.query_sql_text,
    rs.avg_cpu_time / 1000 AS avg_cpu_ms,
    rs.count_executions,
    rs.avg_cpu_time * rs.count_executions / 1000 AS total_cpu_ms
FROM sys.query_store_query q
JOIN sys.query_store_query_text qt ON q.query_text_id = qt.query_text_id
JOIN sys.query_store_plan p ON q.query_id = p.query_id
JOIN sys.query_store_runtime_stats rs ON p.plan_id = rs.plan_id
JOIN sys.query_store_runtime_stats_interval rsi ON rs.runtime_stats_interval_id = rsi.runtime_stats_interval_id
WHERE rsi.start_time >= DATEADD(hour, -1, GETUTCDATE())
ORDER BY rs.avg_cpu_time * rs.count_executions DESC;

Automatic Tuning

Enable Automatic Tuning

-- Enable all auto-tuning options
ALTER DATABASE current
SET AUTOMATIC_TUNING (
    FORCE_LAST_GOOD_PLAN = ON,
    CREATE_INDEX = ON,
    DROP_INDEX = ON
);

-- Check current settings
SELECT * FROM sys.database_automatic_tuning_options;

View Tuning Recommendations

-- Current recommendations
SELECT
    name,
    reason,
    score,
    state_desc,
    is_revertable_action,
    is_executable_action,
    details
FROM sys.dm_db_tuning_recommendations;

Apply Recommendations

-- Force a specific query plan
EXEC sp_query_store_force_plan @query_id = 12345, @plan_id = 67890;

-- Unforce plan
EXEC sp_query_store_unforce_plan @query_id = 12345, @plan_id = 67890;

Hyperscale Features

Storage Auto-Scaling

• Automatically grows up to 128 TB
• No need to pre-provision storage
• Pay only for storage used

Read Scale-Out

-- Connection string option
ApplicationIntent=ReadOnly

-- In application code
"Server=myserver.database.windows.net;Database=mydb;ApplicationIntent=ReadOnly;..."

Named Replicas

-- Create named replica
ALTER DATABASE MyDatabase
ADD SECONDARY ON SERVER MySecondaryServer
WITH (SERVICE_OBJECTIVE = 'HS_Gen5_2', SECONDARY_TYPE = Named, NAME = N'MyReadReplica');

Serverless Configuration

Configure Auto-Pause

-- Via Azure Portal, CLI, or PowerShell
-- Set auto-pause delay (minutes), min/max vCores

-- Check current usage
SELECT
    cpu_percent,
    auto_pause_delay_in_minutes_configured
FROM sys.dm_db_resource_stats_serverless;

Serverless Best Practices

1. Use for intermittent workloads - Saves cost during idle periods
2. Set appropriate min vCores - Prevents cold starts for time-sensitive apps
3. Monitor auto-pause - Auto-resume adds latency
4. Consider always-on for consistent workloads - Provisioned may be cheaper

Connection Optimization

Connection Pooling

// .NET connection string
"Server=tcp:myserver.database.windows.net,1433;Database=mydb;
 Min Pool Size=10;Max Pool Size=100;Connection Timeout=30;"

Retry Logic

// Azure SQL requires retry logic for transient faults
var options = new SqlRetryLogicOption()
{
    NumberOfTries = 5,
    DeltaTime = TimeSpan.FromSeconds(1),
    MaxTimeInterval = TimeSpan.FromSeconds(30)
};

Connection Best Practices

1. Use connection pooling - Reduce connection overhead
2. Implement retry logic - Handle transient faults (error 40613, 40197)
3. Use redirect connection mode - Better performance after initial connection
4. Close connections promptly - Don't hold connections unnecessarily

Azure-Specific Limitations

Not Supported

• SQL Agent (use Azure Functions, Logic Apps)
• BULK INSERT from files (use Blob Storage)
• Linked servers (use Elastic Query)
• FILESTREAM
• Cross-database queries in same server (use Elastic Query)

Workarounds

Bulk Insert from Blob Storage

-- Create credential
CREATE DATABASE SCOPED CREDENTIAL BlobCredential
WITH IDENTITY = 'SHARED ACCESS SIGNATURE',
SECRET = 'your_sas_token';

-- Create external data source
CREATE EXTERNAL DATA SOURCE BlobStorage
WITH (
    TYPE = BLOB_STORAGE,
    LOCATION = 'https://youraccount.blob.core.windows.net/container',
    CREDENTIAL = BlobCredential
);

-- Bulk insert
BULK INSERT MyTable
FROM 'data.csv'
WITH (DATA_SOURCE = 'BlobStorage', FORMAT = 'CSV', FIRSTROW = 2);

Elastic Query for Cross-Database

-- On target database
CREATE MASTER KEY ENCRYPTION BY PASSWORD = 'password';

CREATE DATABASE SCOPED CREDENTIAL ElasticCredential
WITH IDENTITY = 'username', SECRET = 'password';

CREATE EXTERNAL DATA SOURCE RemoteDB
WITH (
    TYPE = RDBMS,
    LOCATION = 'remote-server.database.windows.net',
    DATABASE_NAME = 'RemoteDatabase',
    CREDENTIAL = ElasticCredential
);

CREATE EXTERNAL TABLE dbo.RemoteTable (...)
WITH (DATA_SOURCE = RemoteDB);

Cost Optimization

Reserved Capacity

• Up to 72% savings vs pay-as-you-go
• 1-year or 3-year terms
• Exchange/refund flexibility

Right-Sizing

-- Check if over-provisioned
SELECT
    AVG(avg_cpu_percent) AS avg_cpu,
    MAX(avg_cpu_percent) AS max_cpu,
    AVG(avg_data_io_percent) AS avg_io,
    MAX(avg_data_io_percent) AS max_io
FROM sys.dm_db_resource_stats
WHERE end_time >= DATEADD(day, -7, GETUTCDATE());

-- If avg < 40% consistently, consider downsizing

Hyperscale Cost Considerations

• Compute: Per-second billing
• Storage: Per-hour billing for used space
• Read replicas: Additional compute cost
• Memory not automatically released (monitor and scale appropriately)