Bitvavo – Scaling a 24/7 crypto exchange with Aeron Cluster

At Adaptive’s recent Aeron Community MeetUp in London, Marcos Maia, Senior Staff Engineer, and Oleg Lobanov, Staff Engineer at Bitvavo, described how they rebuilt the core of Bitvavo’s 24/7 crypto exchange on Aeron Cluster. Their talk focused on running Aeron Cluster under real production conditions: keeping the system deterministic and available while processing live orders and maintaining customer balances continuously. Watch the full recording of this session here.

Who is Bitvavo and what do they do?

Bitvavo operates in the European crypto market, a roughly €2.1 trillion market that is larger than Asia and close to North America in size.

The company started in the Netherlands and is now the best-known crypto exchange in the Benelux region, serving customers in the Netherlands, Belgium and Luxembourg.

Over a few years Bitvavo grew from a small startup into a scale-up with more than one million customers, over 500 employees and more than €7 billion in assets under custody. This growth has taken place under rigorous European regulation, which emphasises consumer protection and long-term resilience.

Challenges faced by Bitvavo in rapidly scaling its crypto exchange

Initially, Bitvavo’s platform had been built in startup and MVP mode, which worked well during its earlier growth phase. However, as the business scaled rapidly, expanded across Europe, and began operating in a 24/7 crypto market, Bitvavo needed to redesign core parts of its exchange infrastructure to support higher volumes, new products, and long-term growth.

They required a system that was:

• Always correct and provided deterministic consistency across core exchange services.
• High-performance, with low latency and predictable behaviour under heavy load.
• Capable of running 24/7 with high availability and fault tolerance.
• Able to scale with the business and give Bitvavo more control over future product development and platform optimisation.

Why Bitvavo built on Aeron Cluster

Aeron Cluster was chosen as the foundation for Bitvavo’s core exchange services for several key reasons:

• Performance: Aeron gave Bitvavo the high-performance backbone it needed for services such as the matching engine, OMS, and balances, with predictable ultra-low latency even under heavy load.
• High availability: Aeron Cluster provided the fault tolerance and resilience needed to run critical exchange infrastructure in a 24/7 crypto market with minimal operational disruption.
• Deterministic execution: Aeron Cluster’s replicated state machine model gave Bitvavo deterministic processing across core exchange services, helping the team maintain consistency and making testing and troubleshooting more manageable.

How Bitvavo uses Aeron

Bitvavo uses Aeron Transport, Cluster, and Archive across the core of its exchange platform. Aeron Cluster forms the backbone for core services including the matching engine, OMS, and balances, while Aeron Archive supports downstream data distribution and recovery workflows.

Bitvavo’s exchange architecture includes:

• Gateways: API gateways connect client requests into the cluster for core exchange processing.
• Core clustered services: The cluster is the foundation for Bitvavo’s core services, including the matching engine, OMS, and balance service. Cluster nodes process messages in the same order, apply state changes, and maintain deterministic state across the system.
• Archive: Output data from the cluster is written to Aeron Archive, creating a persistent stream of exchange events.
• Archive tailers: Downstream services tail the archive to produce market data, audit logs, and user transaction history outside the cluster itself.

This architecture gives Bitvavo a deterministic core for critical exchange processing while allowing downstream services to consume and reconstruct exchange outputs in a modular way.

Technical elements and key learnings from Bitvavo’s implementation

The technical implementation of Bitvavo’s exchange platform includes several critical elements that ensure its robustness and efficiency:

• Application-level determinism: Bitvavo learned that determinism must be enforced in the application code, with care taken to avoid non-deterministic behaviour such as HashMap or HashSet iteration, system time usage, randomness, and external side effects inside the state machine logic.
• Layered cluster node divergence protection: Bitvavo built a multi-layer protection model around CI backtesting, standby-node rollout, and continuous live comparison of egress, archive state, and snapshots to detect divergence early and roll back safely if needed.
• Snapshot-based recovery: To recover failed nodes cleanly, Bitvavo stores the last known archive position inside the snapshot, then uses that value to seek the local archive before replay continues, aligning both logical state and physical archive position.
• Aeron Cluster and Aeron Archive: Aeron Cluster provides the replicated, deterministic core for exchange processing, while Aeron Archive records output events that downstream services use for market data, audit logs, and user transaction history.

Recovering from node failure: a new replica joins the cluster from snapshots but starts archiving at position 0, creating offset mismatches for clients tailing the archive.

In production, Bitvavo’s balance and matching engines had been running on Aeron Cluster for 18 months with no issues at the time of the service review and were described internally as the most stable systems in the platform. The exchange was processing typical trading rates of around 30,000 TPS with headroom to 100,000 TPS, up from around 5,000 TPS on the old system, while the next optimization focus remained long-tail latency through legacy infrastructure.

1. 

   Oleg Lobanov Staff Engineer
   Bitvavo

   LinkedIn profile

   Oleg Lobanov is a Staff Engineer at Bitvavo, Europe’s premier Euro-denominated cryptocurrency exchange. He leads the architectural vision and technical evolution of Bitvavo’s ultra-low-latency exchange platform, drawing upon over a decade of experience building distributed, fault-tolerant, and latency-critical systems. Oleg specializes in advanced performance techniques, including JVM garbage collection minimization, optimized data structures, and efficient inter-process communication.

2. 

   Marcos Maia Senior Staff Engineer
   Bitvavo

   LinkedIn profile

   Marcos Maia is a seasoned Senior Staff Engineer at Bitvavo, the leading crypto exchange in the Benelux, with over two decades in the IT industry. A passionate advocate for technology, distributed systems, and engineering culture, Marcos has built a diverse career spanning multiple countries and roles, from engineering to architecture and consultancy. With a focus on finance, crypto, trading, investments, and telecom sectors, he shares his expertise in development practices, low-latency systems, engineering culture, and Apache Kafka through engaging talks for leadership, developers, students, and clients.