Sql-server – sql server spatial index performance

spatialsql server

I have a table with around 2 million records in it. I created a spatial index, using the defaults other than the bounding box. I've been noticing that some queries are extremely fast, and some are extremely slow. The determining factor appears to the size of the polygon used in the query.

On larger search areas, using WITH(INDEX(SIX_FT5)) slows down the query considerably (from 0 seconds, to 15+ seconds). On smaller search areas, the exact opposite it true.

Here's some of the queries I'm testing with:

Fast:

SELECT TOP(1000) 
  * 
FROM [FT5] 
WHERE 
  shape.STIntersects(geometry::STGeomFromText('POLYGON ((-133462.805381701 -668610.241000959, 2934415.68824241 -668610.241000959, 2934415.68824241 2200521.65831815, -133462.805381701 2200521.65831815, -133462.805381701 -668610.241000959))', 2264)) = 1

Slow:

SELECT TOP(1000) 
  * 
FROM [FT5] WITH(INDEX(SIX_FT5)) -- Index hint is the only difference
WHERE 
  shape.STIntersects(geometry::STGeomFromText('POLYGON ((-133462.805381701 -668610.241000959, 2934415.68824241 -668610.241000959, 2934415.68824241 2200521.65831815, -133462.805381701 2200521.65831815, -133462.805381701 -668610.241000959))', 2264)) = 1

Why is that WITH clause slowing things down sometimes?

Best Answer

That your indexing is slowing some queries but not others has a lot to do with the way spatial indexes work:

The spatial index is broken into 4 levels which are broken into grids. With each level you will be able to specify a density which of low, medium or high corresponding to 16, 64 or 256 cells respectively. Then within each cell these are then subdivided into the next lower level with its setting.

All of this means that depending on your settings and your data distribution you may need to adjust the density at different levels to get proper index performance. This may also mean that you need to balance your performance of your main queries of this data against your edge cases.

Take a look at this Microsoft TechNet library article for a more detailed explanation:

Spatial Indexing Overview

Related Solutions

MySQL Syntax – Searching Points Within Polygons Using GeomFromText

I may be wrong, but I believe that MBRWithin(GeomFromText(geo),@zone), should be MBRWithin(geo, @zone) -no need to convert anything, it is stored already in spatial format.

When functions like these get bad-type operands, they return null.

Sql-server – SQL Server spatial filter with additional conditions performance issue

The answer is yes it is possible to get performance with filters over multiple columns including spatial columns.

I've tried to replicate your tables and closely as I can to your brief description using RAND to create prices, location points and link them to events. But I can't duplicate you performance problem. Have I got the basic table details right?

--  Event table
SELECT TOP 30000000 
    N EventID, 
    RAND(CAST(NEWID() AS VARBINARY)) * 10000 + 5000 Price
    into Event
FROM Tally;

ALTER TABLE Event ALTER COLUMN EventID INT NOT NULL;
GO
ALTER TABLE Event ADD CONSTRAINT Event_PK PRIMARY KEY (EventID);
GO
CREATE INDEX Event_Price_IDX ON Event(Price);
GO

-- Location Table
SELECT TOP 35000000
    N LocationID, 
    CEILING(RAND(CAST(NEWID() AS VARBINARY)) * 30000000) EventID,
    Geometry::Point(RAND(CAST(NEWID() AS VARBINARY)) * 710000 + 40000, RAND(CAST(NEWID() AS VARBINARY)) * 710000 + 40000, 0) Location
    into Location
FROM Tally;

Running the 3 versions of your query with hopefully sensible parameters

-- Test Query
DECLARE @price FLOAT = 13000.00 -- ~20% of records
-- 10 sq km query window
DECLARE @BoundingBox Geometry = Geometry::STGeomFromText('POLYGON(( 300000 300000, 310000 300000, 310000 310000, 300000 310000, 300000 300000))',0)

SELECT TOP 2000 [E].*
FROM [dbo].[Event] AS [E]
INNER JOIN [dbo].[Location] AS [L] ON [E].[EventId] = [L].[EventId]
WHERE 
    --[E].[Price] >= @Price
    --AND
    @BoundingBox.STContains([L].[Location]) = 1

SELECT TOP 2000 [E].*
FROM [dbo].[Event] AS [E]
INNER JOIN [dbo].[Location] AS [L] ON [E].[EventId] = [L].[EventId]
WHERE 
    [E].[Price] >= @Price
    --AND
    --@BoundingBox.STContains([L].[Location]) = 1

SELECT TOP 2000 [E].*
FROM [dbo].[Event] AS [E]
INNER JOIN [dbo].[Location] AS [L] ON [E].[EventId] = [L].[EventId]
WHERE 
    [E].[Price] >= @Price
    AND
    @BoundingBox.STContains([L].[Location]) = 1

I get 93ms, 4ms and 93ms respectively, removing the TOP 2000 from the queries causes the price filtered query to blow out to a minute or so.

Are you able to give more details (execution plans, DDL, etc) to help us replicate your issue. My initial thought was that the TOP 2000 was hiding the underlying problem. I think that may still be the case.

Best Answer

Related Solutions

MySQL Syntax – Searching Points Within Polygons Using GeomFromText

Sql-server – SQL Server spatial filter with additional conditions performance issue

Related Question