Facebook Presto

Presto is the distributed DBMS. It is implemented on Java.

A system supports standard language of requests of SQL and allows creation of clusters from hundreds of nodes, processing on them petabytes of data.

The architecture of Presto the words of one of^[1]:

"The client sends SQL to the coordinator of Presto. The coordinator sorts a request, analyzes it, and then plans execution of a request. The manager connects the execution pipeline, gives instructions to nodes which are located most closer to data, and monitors processing promotion. The client takes data from an external step which, in turn, borrows them from lower levels".

Architecture of Presto, 2014

A system is capable to compile requests of SQL in the Java byte code and does it so that to avoid problems with memory allocation and garbage collection. On it transformations of a request on come to an end: the virtual machine performing a bytecode "on the fly" compiles it in machine code. As a result it is executed even quicker.

At the beginning of 2013 the first versions of a system implemented in Facebook. In the spring of 2013 the social network began full-scale transition to Presto. For July 7, 2015 a system works at several huge clusters (the quantity of nodes in one of them can reach thousands), daily solving more than 30 thousand queries to information petabyte.

Source codes of Presto are published on Github service - a repository.

A system is available according to the license Apache.

2016: Teradata actively supported the Presto project

On June 30, 2016 the Teradata corporation announced certification of several solutions for a business intelligence and data visualization in the distributed Presto DBMS.

Information Builders, Looker Data Platform, the platform of visual analytics Qlik, a set of the analytical Tableau and ZoomData tools treat these solutions. The MicroStrategy company undertook to execute certification and completes application testing of Microsoft Power BI.

Several certifications give to customers great opportunities for the choice and promote effective use of professional skills and investments into business intelligence tools to analyze data in Hadoop.

Certification of applications of a business intelligence and analytical applications for use from the distributed Presto DBMS supported by Teradata is important undertaking in the industry. It is very timely step, considering many problems which the companies face at deployment and use of the Hadoop platform now. When the Presto technology, thanks to Teradata corporation, became the solution on accomplishment of SQL queries on Hadoop in the corporate sector, integration into tools of a business intelligence will add necessary functionality for effective use within the organization. Irrespective of the supplier of applications of a business intelligence and analytical applications preferred by the company, the positive result consists in emergence of an effective execution engine of analytical requests which opens the new bright page in the field of visual analytics for all users of Hadoop. Klaudia Imkhoff (Claudia Imhoff), leading consultant for a business intelligence and data warehouses, founder of Boulder BI Brain Trust fund

For June, 2016 Presto allows to send requests in the different file systems supported by the Hadoop platform, including HDFS, Amazon S3, Cassandra, relational databases and even corporate data warehouses – and is suitable for analysts of data which requests require a response within seconds or minute.

Consolidation of the distributed Presto DBMS supported by Teradata with several tools of a business intelligence means that the companies can develop applications and reports of a business intelligence on platforms, using the distributed Presto DBMS, the ODBC and JDBC drivers for Presto from Teradata. Such integration helps to provide rapid application development and the major data on analysis results by the client of intensity of events, the analysis of outflow of clients, information analysis from sensors for visualization of Internet of Things and many other things. As a result, the companies can quicker create for themselves new opportunities, effectively using the available investment resources in based on SQL applications for a business intelligence and Hadoop.

Notes