Range Index in Apache Pinot

The Range Index in Apache Pinot divides numeric values into predefined ranges (buckets). These ranges are created during segment creation, and helps Pinot quickly filter queries like price > 500 or age BETWEEN 20 AND 30 kinds of queries efficiently.

 

Let’s take a sample dataset:

product_id	name	price	timestamp
1	Laptop	1000	1700000000
2	Phone	700	1700005000
3	Tablet	500	1700010000
4	Keyboard	50	1700020000
5	Monitor	300	1700030000

 

Let's a Range Index on:

·      price (INT)

·      timestamp (LONG)

 

Range Index for price

Range	Records in this range
0-250	[]
251-500	[3, 5]
501-750	[2]
751-1000	[1]

 

Range Index for timestamp

Range	Records in this range
1699999000-1700005000	[1, 2]
1700005001-1700015000	[3]
1700015001-1700025000	[4]
1700025001-1700035000	[5]

 

For example, when you execute following query

SELECT * FROM products WHERE price > 500;

 

How Pinot Uses the Range Index?

·      Pinot finds relevant ranges:

o   It skips 0-250 and 251-500 ranges.

o   It directly looks up 501-750 and 751-1000, and get the records {2, 1}

·      It retrieves rows 2 and 1 (Phone and Laptop).

·      No full table scan is needed in this case

 

In summary,

·      Range Indexes are optimized for BETWEEN, >, < queries.

·      Best for analytics on numeric/timestamp columns.

 

Previous                                                    Next                                                    Home