Casumo is a technology startup formed in 2012, developing an award-winning online casino. Casumo’s concept is changing the common perception of casinos. Through design, technology, and innovation, and by ignoring preconceptions and rules, Casumo aims to disrupt a young and in many ways immature, industry.

From idea to

It all started out with a group of tech guys who worked at an online casino and who saw huge potential in developing a technologically unique casino system. They had fun together and were motivated to learn and implement the newest technological concepts to build an advanced system which would power the big online casino games in the Nordic countries following with UK and other countries as well. An important source of their inspiration was Greg Young and his talks on CQRS and Event Sourcing. That convinced them that CQRS-based architecture was the best way to develop the new casino platform. The goal of the team was not just to build a scalable casino platform but to build a scalable development team as well.

“To be more confident when developing a system based on CQRS we chose Axon Framework because it was the most complete framework which enabled the CQRS implementations.”

After spending some time researching the implementation and benefits of CQRS, they finally settled on architecture based on CQRS and Microservices. The architecture’s loosely coupled components were particularly appealing to the team, as the system could then be divided into smaller sections and distributed to smaller teams. The teams worked in parallel without the need to oversee the whole system, resulting in faster development of the small, isolated software components. The Casumo team had the guts to choose some not yet very commonly used CQRS principles and, after some research, they discovered that Axon is the only complete framework that enables CQRS implementation. Building their own framework was not an option, as that would take time and distract them from their main business focus. They decided to use Axon Framework, thereby moving the project from paper to reality. In addition to the architecture requirements and distribution of work to small, scalable teams, the modern agile development method was a fixed prerequisite. CQRS does not itself provide an agility aspect, but it forces the development mindset to simplify loosely coupling different systems.

Challenges: Keep the spin button FAST




When doing something new in a relatively young, highly competitive industry, distinguishing yourself is important. The right technology is essential to achieving that. The team’s motto is “Keep that spin button fast” – slot machines require instant feedback, as every millisecond of delay translates to lost revenue. The project development requirements:

  • High performance in a high-transaction environment;
  • Complex requirements such as CQRS-based design, highly efficient system design, scalable development teams, scalable deployment, etc.;
  • Agile development method;
  • Short time to market, develop the system as quickly and as stable as possible.

At the system’s core is the online wallet that keeps track of all casino transactions. It generates a large amount of data, which requires good, scalable storage. A lot of thought went into the infrastructure design to optimally support data creation, storage, and backup. The system design needed to be scalable for a huge number of transactions and store billions of events generated by online wallet. An adverse side effect of the large amount of data is that it becomes nearly impossible to search.

“That service should have its own query model.…“

Although a lot of attention went into the wallet’s design, it eventually required refactoring for improved cohesion and performance. During the wallet redesign process, Percona database was introduced as an event store, aiding the optimization of the MySQL storage tremendously. The event store design also underwent a shift during the development process. At the start of the project, a common event store was used for all the aggregates, and sharing took place at the application level. Moving forward with the Microservices architecture, each service receives its own event store database. Most of them still reside on the same server, but they can easily be relocated for load handling if necessary.

Other parts of the system

The team experimented with many different technologies and solutions, focusing on keeping things simple wherever possible. For example:

  • Two different flavors of distributed caches, Hazelcast and Terracotta, were used for load handling. After some experimentation, Hazelcast was chosen and is still in use for some of the services. The use of local Guava caches is currently preferred, as they make removal of any cross service dependencies easier. As they only need an event handle that builds the cache for you, they are also very easy to populate in an event-sourced system.
  • As the Event store was the continuous focus, the team also tried various NoSQL databases. Percona and its support were the most comfortable and are used now.

Solution using Axon Framework

Casumo’s design was based on the business needs, meaning that everyone in the business is able to understand what every event means; the platform is modeled according to the business language. In this design, CQRS definitely helped steer the project to the right design and Axon helped with modeling and solving common, nonfunctional problems. CQRS and event sourcing forced the construction of an understandable, scalable, and maintainable system. mainly uses Polyglot Microservices architecture with Axon features for the command bus, event sourcing, and Sagas (the building blocks of CQRS). The framework was used pragmatically where needed and provided good input throughout the project, particularly for the infrastructure and data processing issues. Experimenting with new things required expert knowledge. The modeling and ironing out of performance issues benefited greatly from the expert consultation. The Casumo development team worked closely together with the Axon team for modeling and performance issues, and with the Percona* team for leveraging MySQL for the event stores as much as possible.

In summary

Microservices were used to help establish the scalable teams by mapping the problems to smaller chunks which could then be independently maintained by the teams. Finding the correct boundaries took some trial-and-error, with larger services eventually being split up and some smaller services merged during the development process.

Event sourcing is ensuring that every change to the state of an application is captured in an event object, and that these event objects are themselves stored in the sequence they were applied for the same lifetime as the application state itself. With out-of-the-box solutions, the system became unstable and crashed sometimes, requiring a lot of attention along the way.

Event based integration automatically leads to the correct domain model when event sourcing is applied, as long as the event flow is properly understood.

Lessons learned: when shared queries cause services to become too tightly coupled, they must be de-coupled. Keep the query model isolated, sharing as little as possible, especially not the database.

“Axon Framework is aimed at making life easier for developers who want to create a java application based on the CQRS principles. It improved our development speed and made us more comfortable with the system. It also forced us to apply a number of methods which have benefited us going forward.”

*Percona is an open source software company specialized in MySQL Support, Consulting, Managed Services, and Training