Partizionamento hash per la distribuzione

Welcome to Hash Partitioning! This strategy helps distribute data evenly across multiple partitions. Instead of dividing data by ranges or specific values, it uses a hash function.

This approach is especially useful when you want to avoid 'hot spots' where one partition receives significantly more data or queries than others.

Hash partitioning is ideal for:

• Even Distribution: It spreads data uniformly across partitions, preventing any single partition from becoming too large or slow.
• Avoiding Hot Spots: Useful when your partition key doesn't have a natural order or predictable distribution (unlike time-series data for Range Partitioning).
• Improved Concurrency: Queries can often operate on different partitions in parallel, boosting performance.

To use hash partitioning, you define a parent table with PARTITION BY HASH (column_name). Then, you create child partitions using FOR VALUES WITH (MODULUS n, REMAINDER r).

• MODULUS (n): The total number of partitions you plan to have.
• REMAINDER (r): A unique number from 0 to n-1, indicating which 'bucket' of hashed values this partition will hold.

Let's create a table for user activity logs, partitioned by a user_id. We'll start with a parent table and two partitions. Note how the REMAINDER values cover the MODULUS.

When you insert data, PostgreSQL calculates a hash value for the user_id. It then performs a modulo operation (hash_value % MODULUS). The result determines which partition the row belongs to.

For example, if MODULUS is 2:

• If hash_value % 2 = 0, it goes to user_activity_0.
• If hash_value % 2 = 1, it goes to user_activity_1.

Let's insert some data into our user_activity table. PostgreSQL automatically directs each row to the correct hash partition based on the user_id.

Try inserting different user_id values and observe the distribution!

Querying a hash-partitioned table is just like querying a regular table. PostgreSQL's query planner is smart enough to use partition pruning.

If your query includes the partition key (e.g., WHERE user_id = 101), the planner will only scan the relevant partition, significantly speeding up queries.

Unlike Range partitioning (based on intervals) or List partitioning (based on specific values), Hash partitioning doesn't group data logically.

Its strength is in spreading data evenly, making it great for keys without natural ranges (like UUIDs or arbitrary IDs) and for preventing specific parts of your data from becoming bottlenecks.

When using hash partitioning:

• Choose your key wisely: The partition key should have a good distribution of hash values.
• Modulus & Remainder: Ensure all REMAINDER values from 0 to MODULUS - 1 are covered by partitions.
• Adding/Removing Partitions: Changing the MODULUS later requires re-hashing all data, which can be complex. Plan your partition count carefully.

Which scenario is best suited for PostgreSQL hash partitioning?

In this lesson, you've learned about PostgreSQL hash partitioning:

• It distributes data evenly based on a hash function of the partition key.
• It uses MODULUS and REMAINDER to define partitions.
• It's excellent for avoiding hot spots and improving performance for non-sequential or arbitrary keys.
• Queries benefit from partition pruning when the partition key is used.

Hash partitioning is a powerful tool for scaling your database when even data spread is crucial.