Apache Pulsar Outperforms Apache Kafka by 2.5x on OpenMessaging Benchmark!

Open messaging

In a recent announcement by industry analyst firm Gigaom, the OpenMessaging benchmark performance results showed that Apache Pulsar delivers consistently superior throughput and latency even if at an increasing scale. OpenMessaging is a collaborative Linux Foundation effort that is supported by many large-scale companies with a mission to create a vendor-neutral, globally adopted and open standard for distributed messaging that can be deployed in on-premise, cloud and hybrid use cases. According to Gigaom, Apache Pulsar has outpaced Kafka cross all the workloads tested in their evaluation using the OpenMessaging benchmark.

Here are the key results of the evaluation of Apache Pulsar and Apache Kafka using OpenMessaging benchmark:

· Around 150% higher maximum throughput with Apache Pulsar.

· Around 40% lower message latency and greater consistency in latency.

· Better scalability that delivered consistent results across a range of message sizes and partition counts.

Let us discuss some points why Apache Pulsar is better than Kafka to build messaging service.

1. Streaming and queuing comes together: Apache Pulsar serves two purposes together. It handles high-rate, real time use cases (like Kafka) but also supports standard message queuing patterns like fail-over subscriptions, competing consumers and more. As it can do two things in one, Pulsar is always a good deal.

2. Partitions: In Kafka, topics are partitioned which is important for increasing throughput. By distributing the work through partition into multiple brokers, the processing rate by single topic can be increased. But there could be cases when a user only needs a single topic. Pulsar subscriptions allow users to add as many consumers as they want in one topic. However, it also offer partitioned topic if the user needs them.

3. Logs: Kafka offers an insight in logs in which data can be easily written and extracted. But, there could be cases where data is large and adjusting a log in a single server could be challenging. Apache Pulsar breaks the logs into segments and distributes these segments into multiple servers.

4. Stateless brokers: Stateless components start quickly; they are interchangeable and scale seamlessly. Kafka brokers are not stateless. In them, if a broker fails, other broker can take over it and the user will not be able to add another broker. On the other hand, in Apache Pulsar architecture, brokers are stateless. They will be able to persist messages. So Apache Pulsar maintains state, and not only brokers. So, in case of Pulsar, if the loads get high, the user only has to add another broker and can get the work done.

5. Geo-replication for dummies: Geo-replication is an important feature in Pulsar. Pulsar was designed with keeping geo-replication in mind. Configuring it is easy and works.

6. Consistently Faster: Benchmark testing has proved that Pulsar delivers higher throughput along with lower and more consistent latency.

Read More… http://entradasoft.com/blogs/apache-pulsar-outperforms-apache-kafka-on-openMessaging-benchmark