Agent Skills: Index Strategies

Index Strategies

Comprehensive guide to SQL Server index design and optimization.

Quick Reference

Index Types

| Type | Description | Best For | |------|-------------|----------| | Clustered | Table data order | Primary access path, range scans | | Nonclustered | Separate structure | Specific query patterns | | Columnstore | Column-based storage | Analytics, aggregations | | Filtered | Partial index | Well-known subsets | | Covering | All columns needed | Avoiding key lookups |

Clustered Index Guidelines

Ideal Clustered Key:

• Narrow (small data type)
• Unique or mostly unique
• Ever-increasing (identity, sequential GUID)
• Static (rarely updated)

-- Good: Identity column
CREATE CLUSTERED INDEX CIX_Orders ON Orders(OrderID);

-- Good: Sequential GUID
CREATE TABLE Orders (
    OrderID UNIQUEIDENTIFIER DEFAULT NEWSEQUENTIALID() PRIMARY KEY CLUSTERED
);

-- Avoid: Wide composite keys, frequently updated columns, GUIDs (NEWID)

Nonclustered Index Design

-- Basic index
CREATE NONCLUSTERED INDEX IX_Orders_CustomerID
ON Orders(CustomerID);

-- Covering index (avoids key lookup)
CREATE NONCLUSTERED INDEX IX_Orders_CustomerID_Cover
ON Orders(CustomerID)
INCLUDE (OrderDate, TotalAmount, Status);

-- Filtered index (partial)
CREATE NONCLUSTERED INDEX IX_Orders_Active
ON Orders(CustomerID, OrderDate)
WHERE Status = 'Active';

-- Descending order
CREATE NONCLUSTERED INDEX IX_Orders_DateDesc
ON Orders(OrderDate DESC, OrderID DESC);

Index Selection Guide

By Query Pattern

| Pattern | Recommended Index | |---------|-------------------| | WHERE Col = value | Nonclustered on Col | | WHERE Col = v1 AND Col2 = v2 | Nonclustered on (Col, Col2) | | WHERE Col = v ORDER BY Col2 | Nonclustered on (Col, Col2) | | WHERE Col BETWEEN x AND y | Col as leftmost key | | SELECT * WHERE Col = v | Clustered or covering NC | | Large aggregations | Columnstore | | Specific subset queries | Filtered index |

Column Order in Composite Keys

-- Order matters! Left-to-right matching
CREATE INDEX IX_Example ON Table(A, B, C);

-- These queries CAN use the index:
WHERE A = 1
WHERE A = 1 AND B = 2
WHERE A = 1 AND B = 2 AND C = 3
WHERE A = 1 AND B > 5 ORDER BY B

-- These queries CANNOT use index seek:
WHERE B = 2                    -- A not specified
WHERE B = 2 AND C = 3          -- A not specified
WHERE A = 1 AND C = 3          -- B skipped (partial match only)

Columnstore Indexes

Clustered Columnstore

-- Best for data warehousing
CREATE CLUSTERED COLUMNSTORE INDEX CCI_FactSales
ON FactSales;

-- Ordered columnstore (SQL 2022+)
CREATE CLUSTERED COLUMNSTORE INDEX CCI_FactSales
ON FactSales
ORDER (DateKey, ProductKey);

Nonclustered Columnstore

-- Hybrid OLTP/OLAP
CREATE NONCLUSTERED COLUMNSTORE INDEX NCCI_Orders_Analysis
ON Orders(OrderDate, ProductID, Quantity, Amount)
WHERE Status = 'Completed';

Columnstore Best Practices

1. Load batches >= 102,400 rows - Creates compressed segments
2. Order data by filtered columns - Better segment elimination
3. Use REORGANIZE, not REBUILD - More efficient maintenance
4. Avoid frequent small updates - Causes deltastore fragmentation
5. Partition by date - Enables partition elimination

-- Maintenance
ALTER INDEX CCI_FactSales ON FactSales REORGANIZE;

-- Check fragmentation
SELECT
    object_name(object_id) AS TableName,
    index_id,
    avg_fragmentation_in_percent,
    fragment_count
FROM sys.dm_db_index_physical_stats(DB_ID(), NULL, NULL, NULL, 'LIMITED');

Filtered Indexes

-- Index active orders only
CREATE NONCLUSTERED INDEX IX_Orders_Active
ON Orders(CustomerID, OrderDate)
WHERE Status = 'Active';

-- Index non-NULL values
CREATE UNIQUE INDEX IX_Users_Email
ON Users(Email)
WHERE Email IS NOT NULL;

-- Constraints:
-- - Cannot use variable in filter
-- - Query WHERE must match or be subset of filter WHERE
-- - May cause parameter sniffing issues

Covering Indexes

-- Eliminate key lookups
-- Original: Index on CustomerID, query selects OrderDate, Amount
-- Execution plan shows Key Lookup

-- Solution: Covering index
CREATE INDEX IX_Orders_CustomerID_Cover
ON Orders(CustomerID)
INCLUDE (OrderDate, Amount, Status);

-- INCLUDE columns:
-- - Not in key (not sorted)
-- - Stored at leaf level only
-- - Don't contribute to 900-byte key limit
-- - Perfect for frequently selected columns

Index Maintenance

Fragmentation Guidelines

| Fragmentation % | Action | |-----------------|--------| | < 5% | None needed | | 5-30% | REORGANIZE | | > 30% | REBUILD |

-- Reorganize (online, minimal locking)
ALTER INDEX IX_Orders_CustomerID ON Orders REORGANIZE;

-- Rebuild (offline by default, more thorough)
ALTER INDEX IX_Orders_CustomerID ON Orders REBUILD;

-- Online rebuild (Enterprise Edition)
ALTER INDEX IX_Orders_CustomerID ON Orders
REBUILD WITH (ONLINE = ON);

-- Resumable rebuild (SQL 2017+)
ALTER INDEX IX_Orders_CustomerID ON Orders
REBUILD WITH (ONLINE = ON, RESUMABLE = ON, MAX_DURATION = 60);

-- Resume interrupted rebuild
ALTER INDEX IX_Orders_CustomerID ON Orders RESUME;

Statistics Update

-- Update after index changes
UPDATE STATISTICS Orders;

-- Full scan for accurate stats
UPDATE STATISTICS Orders WITH FULLSCAN;

-- Check last update
SELECT
    OBJECT_NAME(object_id) AS TableName,
    name AS StatsName,
    STATS_DATE(object_id, stats_id) AS LastUpdated
FROM sys.stats
WHERE object_id = OBJECT_ID('Orders');

Performance Monitoring

Index Usage Stats

SELECT
    OBJECT_NAME(i.object_id) AS TableName,
    i.name AS IndexName,
    ius.user_seeks,
    ius.user_scans,
    ius.user_lookups,
    ius.user_updates
FROM sys.indexes i
LEFT JOIN sys.dm_db_index_usage_stats ius
    ON i.object_id = ius.object_id
    AND i.index_id = ius.index_id
WHERE OBJECTPROPERTY(i.object_id, 'IsUserTable') = 1
ORDER BY ius.user_seeks + ius.user_scans DESC;

Missing Index Recommendations

SELECT
    migs.avg_user_impact AS ImpactPercent,
    mid.statement AS TableName,
    mid.equality_columns,
    mid.inequality_columns,
    mid.included_columns
FROM sys.dm_db_missing_index_groups mig
JOIN sys.dm_db_missing_index_group_stats migs
    ON mig.index_group_handle = migs.group_handle
JOIN sys.dm_db_missing_index_details mid
    ON mig.index_handle = mid.index_handle
ORDER BY migs.avg_user_impact DESC;