Checkpoints

Not Supported

PrestoDB does not support checkpointing. As is the case for many OLAP database management systems, Presto assumes that the nodes will not fail during the execution. The client has to rerun the entire query once failure occurs.

Compression

Dictionary Encoding Run-Length Encoding

PrestoDB can operate on dictionary and run-length-encoded blocks from connectors. When generating intermediate results, Presto also produces compressed data in the form of dictionary or run-length-encoded blocks.

Concurrency Control

Not Supported

PrestoDB is a query engine designed for query large amounts of data, and it does not support updating the data that it queries. As a result, concurrency control is not needed for the usages of PrestoDB.

Data Model

Column Family / Wide-Column Relational Key/Value Document / XML

PrestoDB provides connectors to different data sources, and each connector is implemented to be compatible with the data model of the underlying database. Thus, PrestoDB supports different types of data models, including column family, document/XML, key/value, and relational.

Foreign Keys

Not Supported

Indexes

Not Supported

PrestoDB does not support indexes. All intermediate processing and storage are done in the memory to avoid unnecessary I/O overheads.

Isolation Levels

Read Uncommitted Read Committed Serializable Repeatable Read

Depending on the underlying source of data, whether or not transaction is supported depends on the implementation of the specific connector. For the connectors that support transactions, the PrestoDB API supports four different types of isolation levels. The isolation level is to be specified when a transaction is started.

Joins

Hash Join Broadcast Join Shuffle Join

PrestoDB has two types of join distributions. It can support both broadcast join and partitioned (shuffle) join. The join distribution can either be specified by the user or be decided by the cost-based optimization strategies that are supported by Presto.

At each node level, PrestoDB performs a hash-based join.

Logging

Not Supported

By default, PrestoDB does not support logging. Customized tools can be added to PrestoDB to log the events and queries at the node level.

Query Compilation

Code Generation JIT Compilation

PrestoDB uses code generation targeting JVM bytecode. To do this, it can evaluate expressions, and also use heuristics to generate code that are compatible with the optimization of the JIT compilers, thus providing better performance.

Query Execution

Materialized Model

PrestoDB uses the materialized model and emits the entire result at a time. The result is then stored in the memory and the coordinator decides its usage.

Query Interface

SQL

PrestoDB follows the ANSI SQL specification. To improve usability, PrestoDB also supports anonymous functions and built-in higher-order functions, including transform, filter and reduce.

Storage Architecture

In-Memory

PrestoDB is designed to query data from sources including Hadoop environments and other relational database systems, so it does not directly take the role of data storage. All data and the intermediate results are stored in-memory whenever possible. For communication between nodes, data is also stored in in-memory buffers and sent through the network. This avoids the high cost of I/O operations and speeds up the execution.

For memory-intensive queries, PrestoDB also offers the functionality of spilling data to disk. But this is not a primary function of PrestoDB and it is assumed that most of the query operations should be performed in-memory completely.

Storage Format

Apache Iceberg

Storage Model

Decomposition Storage Model (Columnar)

To execute a query, PrestoDB splits the assignments to each worker, and the workers fetch the data from the data sources. The unit of data that PrestoDB locally operates on is called a page. The page is a columnar of a sequence of rows.

Stored Procedures

Supported

Stored procedures may be provided by the connectors to data sources. PrestoDB provides a "call" statement to invoke a procedure. However, some systems, such as PostgreSQL, have their own stored procedures that are not directly callable by PrestoDB.

System Architecture

Shared-Nothing

As similar to many classic MPP (massively parallel process) database management systems, PrestoDB utilizes a shared-nothing system architecture.

PrestoDB is deployed on a cluster of nodes. A node can take the role of either a coordinator or a worker. Each node has its own private disk and memory, and the user can configure the memory usage of each node. Since PrestoDB does not store data directly, the disk of each node is used minimally for storing logs only, and all communications are done through the network.

Views

Virtual Views

PrestoDB provides functions to store the view as a logical table that can be used by future queries. The view does not contain any data, meaning that the query that generates the view has to be executed when the view is referenced.