Home>Articles>The rise of event stream processing
A business man in business casual outfit holding an iPad and looking at it, background image with virtual data graphs on it
Articles Popular

The rise of event stream processing

“The perfect kind of architecture decision is the one which never has to be made”  Robert C. Martin

From being a phenomenal trend to becoming one of the most essential IT designs considerations, event streaming has evolved to be pivotal for businesses due to an exponential increase in data and growing requirements for them to be real-time.

A customer no longer waits for three hours before receiving a confirmation of his online shopping payment or book his Uber four hours in advance every time he needs to travel somewhere. Businesses need to react in real-time. They need to be event-driven.

What is event stream processing?

Event stream processing is what enables the system to analyse the data about an event and respond to it in real-time. As in the name, there are three parts of it: Event, stream and processing.

Let’s start with events. Event is simply something that happens. It is an action that is generated asynchronously from the external environment and is recognised by the software.

For example, once a customer places an online order, it is an event. This event is followed by other events like generating an invoice, creating a payment mode, sending a confirmation of the payment via email. The faster the system responds, the better the customer experience.

 

 

Example Events

User login

Item selected

Order placed

Payment generated

Confirmation email sent

In a software application, there are two aspects of an event.
First, an Action which is User log in, order made and the Context that provides details like the full name of the user, the order number or the confirmation mail id. Secondly, a stream is the constant flow of data events from various devices connected to your system and processing is the analysing of these events. Therefore, event stream processing is analysing the events as they are generated.

How does event stream processing work?

For traditional data processing, the most accepted way is batch processing in legacy applications. it is the processing of a large volume of data at once. The data comprises millions of records for a day and can be stored in a variety of ways (file, record). The jobs are typically completed simultaneously in a non-stop sequential order.

For example, a financial company may process all its payroll and invoicing related transactions periodically. However, this requires huge data storage (depending on the size of the company) and the response time is delayed.

Event stream processing, on the other hand, gives a contextual real-time view of the data. It processes events as they occur and integrates them with the already stored information to facilitate instant data analysis. It is a continuous computation method and the data flows through the systems with no compulsory time limitations on the output.

In other words, event stream processing analyses high-velocity big data in motion. It enables data filtering, aggregation and cleansing before storing the data. Event processing becomes critical when the events need to be monitored and responded frequently. One of the biggest advantages of event streaming is fraud detection. If the data is event stream-processed, fraudulent transactions can be identified and stopped even before the completion.

Why event stream processing?

Businesses, which do not require high-velocity big data analysis, respond real-time to the changing marketing conditions or scale their operations as the volumes go up, do not need event stream processing. However, the critical question is, which business in today’s time would not want to respond to the market in real-time?
The importance of event stream processing arises from the fact that it overcomes the challenges of:

  • Analysis of high-velocity data in motion
  • Real-time processing of massive volume of events
  • Respond real-time to the customer requirements
  • Continuous data monitoring
  • Scalability
  • Identifying problems/frauds quickly

Some of the popular use cases for event stream processing are produce-consumer API, streaming, custom connectors for data ETL pipeline and log aggregation. Most of the banking applications use the pipeline and streaming features to fast integrate different applications and to prevent fraudulent transactions.
Similarly, retailers like Alibaba and Zalando use event stream processing for business process monitoring and continuous Extract, Transform and Load (ETL), online recommendations etc.

Because of the streaming internal mechanism, we can think about a zero data loss system where the requested data will be retained (configurable period) when the consumer is down for some reason. Also, in the world of Microservices and DevOps, event streaming supports automatic failover, high availability and message replication.

Conclusion

With event stream processing, we can think about an IT architecture that is close to zero data loss and super-fast processing. Event stream processing has garnered a lot of attention from the business community as well as developers. There are multiple forums where individuals are sharing knowledge about event streaming. Companies are organising summits to spread awareness, hands-on training, networking opportunities and to learn more about the event-driven ecosystem, a system that is digital and agile.

Apart from apache, which is providing an open-source version of Kafka for streaming, big names like Confluent and Debezium also launched their customised versions. Companies like Mercedes Benz, Linkedin, Comcast, Data Visor are using event-driven architecture for their IT system.

Written by Swarnava Chakraborty

For similar articles, please read further:

https://www.thoughtworks.com/radar/techniques/event-streaming-as-the-source-of-truth

http://fastnforward.technaura.com/event-streaming-using-apache-kafka/

Making sense of stream processing

Leave a Reply

Your email address will not be published. Required fields are marked *