Apache Pulsar™Apache Pulsar is an open-source, distributed messaging and streaming platform built for the cloud.
Cloud-Native, Distributed Messaging and Streaming
What is Pulsar
Apache Pulsar is an all-in-one messaging and streaming platform. Messages can be consumed and acknowledged individually or consumed as streams with less than 10ms of latency. Its layered architecture allows rapid scaling across hundreds of nodes, without data reshuffling.
Its features include multi-tenancy with resource separation and access control, geo-replication across regions, tiered storage and support for six official client languages. It supports up to one million unique topics and is designed to simplify your application architecture.
Pulsar is a Top 10 Apache Software Foundation project and has a vibrant and passionate community and user base spanning small companies and large enterprises.
Rapid Horizontal ScalabilityScales horizontally to handle the increased load. Its unique design and separate storage layer enable handling the sudden surge in traffic by scaling out in seconds.
Low-latency messaging and streamingAcknowledge messages individually (RabbitMQ style) or cumulative per partition (i.e., offset-like). Enables use cases such as distributed work queues or order-preserving data streams at very large scales (hundreds of nodes) and low latency (<10ms).
Seamless Geo-ReplicationProtect against complete zone outages using replication across different geographic regions. Flexible and configurable replication strategies across distant Pulsar Clusters. Uniquely supports automatic client failover to healthy clusters.
Multi-tenancy as a first-class citizenMaintain one cluster for your entire organization using tenants. Access control across data and actions using tenant policies. Isolate specific brokers to a tenant when maximum noisy neighbor protection is needed.
Automatic Load BalancingAdd or remove nodes and let Pulsar load balance topic bundles automatically. Hot spotted topic bundles are automatically split and evenly distributed across the brokers.
Official multi-language supportOfficially maintained Pulsar Clients for Java, Go, Python, C++, Node.js, and C#.
Official 3rd party integrationsPulsar has officially maintained connectors with popular 3rd parties: MySQL, Elasticsearch, Cassandra, and more. Allows streaming data in (source) or out (sink).
Serverless FunctionsWrite and deploy functions natively using Pulsar Functions. Process messages using Java, Go, or Python without deploying fully-fledged applications. Kubernetes runtime is bundled.
Supports up to 1M topicsPulsar's unique architecture supports up to 1 million topics in a single cluster. Simplify your own architecture by avoiding multiplexing multiple streams into a single topic.
How does Pulsar work
Producer & Consumer
A Pulsar client contains a consumer and a producer. A producer writes messages on a topic. A consumer reads messages from a topic and acknowledges specific messages or all up to a specific message.
Pulsar and BookKeeper use Apache ZooKeeper to save metadata coordinated between nodes, such as a list of ledgers per topic, segments per ledger, and mapping of topic bundles to a broker. It’s a cluster of highly available and replicated servers (usually 3).
Topics (i.e., partitions) are divided among Pulsar brokers. A broker receives messages for a topic and appends them to the topic’s active virtual file (a.k.a ledger), hosted on the Bookkeeper cluster. Brokers read messages from the cache (mostly) or BookKeeper and dispatch them to the consumers. Brokers also receive message acknowledgments and persist them to the BookKeeper cluster as well. Brokers are stateless (don't use/need a disk).
Apache BookKeeper is a cluster of nodes called bookies. Each virtual file (a.k.a ledger) is divided into consecutive segments, and each segment is kept on 3 bookies by default (replicated by the client - i.e., the broker). Operators can add bookies rapidly since no data reshuffling (moving) between them is required. They immediately share the incoming write load.