Overview

LogDevice is a distributed log system. While a file system stores and serves data organized as files, a log system stores and delivers data organized as logs. The log can be viewed as a record-oriented, append-only, and trimmable file. In more detail:

• Record-oriented means that data is written into the log as indivisible records, rather than individual bytes. A record is the smallest unit of addressing: a reader application always starts reading a log from a particular record, or from the next record to be appended to the log. The reader receives one or more records at a time. Records of a log are identified by monotonically increasing log sequence numbers (LSNs). Record numbering is not guaranteed to be continuous. There may be gaps in the numbering sequence. The writer does not know in advance what LSN its record will be assigned upon a successful write.

• Logs are append-only. No support for modifying existing records is provided.

• Logs are expected to live for a relatively long time: days, months, or even years before they are deleted. The primary space reclamation mechanism for logs is trimming: dropping the oldest records according to either a time-based or space-based retention policy, or in response to an explicit request to trim a log.

LogDevice is designed from the ground up to serve many types of logs with high reliability and efficiency at scale. It is also highly tunable, allowing each use case to be optimized for the right set of trade-offs in the durability-efficiency and consistency-availability space. Here are some examples of workloads supported by LogDevice:

• Write-ahead logging for durability
• Transaction logging in a distributed database
• Event logging
• Stream processing
• ML training pipelines
• Replicated state machines
• Journals of deferred work items

Getting started

Start by running a LogDevice cluster locally.

After that, try Creating your first cluster, which tells you how to configure a fully functional LogDevice cluster on multiple servers.

To learn more about LogDevice, there's an explanation of its architecture. If you want to write an application that interacts with a cluster, see the Client library API.

The build guide explains how to obtain the source code and build LogDevice components including the logdeviced server, the client library, and the administrative shell utility called ldshell.

Administration

This release contains minimal support for LogDevice cluster administration. You can use the supplied ldshell utility to create and configure logs. LDQuery provides a powerful SQL-based mechanism for querying the state of a LogDevice cluster and its various components, as well as several types of log metadata.