IPF Processes 6,000 Payments Per Second With Zero Data Loss

Icon Solutions and MongoDB have reported benchmark results showing that the Icon Payments Framework (IPF), running on MongoDB Atlas, can process 6,000 payment transactions per second while maintaining low latency and recovering from infrastructure failures without losing transactions. The results highlight how modern real-time payment infrastructure can support the growing demands of instant payments while maintaining resilience and scalability.

IPF Demonstrates High-Volume Payment Processing

The joint benchmark evaluated IPF and MongoDB Atlas under increasing transaction loads, starting at 500 payments per second and scaling up to 6,000 payments per second.

At the highest tested level, IPF processed 6,000 end-to-end payment transactions every second. The system completed all 12 processing stages associated with the reference payment flow, demonstrating that the result represented full payment processing rather than a simplified database transaction test.

The benchmark generated approximately 430,000 database operations per second across two MongoDB Atlas clusters. Around 236,000 operations were associated with the IPF Event Journal, while approximately 196,000 were handled by the IPF Operational Data Store.

Sub-Second Latency at 6,000 Transactions Per Second

Speed was another important component of the benchmark.

At 6,000 transactions per second, IPF recorded a mean end-to-end latency of 0.51 seconds. Persistence latency remained below 30 milliseconds even at peak throughput.

The results are significant for instant-payment environments, where transactions must often be completed within strict processing windows. According to the benchmark results, the 0.51-second average remained well below the 10-second window typically associated with instant payment schemes.

Maintaining low latency while processing thousands of transactions simultaneously can be challenging for financial institutions, particularly as payment traffic moves increasingly toward real-time rails.

Testing the Complete Payment Journey

The benchmark was designed to replicate a realistic SEPA Instant outbound payment process rather than testing only individual components.

Each transaction passed through 12 processing steps. These included duplicate checking, message validation, scheme-rule validation, fraud and sanctions screening, foreign-exchange retrieval, funds reservation, booking, and payment message generation.

The surrounding banking systems were represented using simulators, allowing the test to focus on the performance and resilience of the IPF and MongoDB Atlas environment.

This approach gave the companies an opportunity to evaluate how the payment platform behaved as transaction volumes increased and infrastructure failures were introduced.

Near-Linear Scaling Across the Benchmark

The benchmark also demonstrated predictable scaling as transaction volumes increased.

IPF and the underlying database environment were tested progressively from 500 to 6,000 transactions per second. According to the results, application and database resources scaled in a near-linear fashion, with per-shard CPU usage and database insert rates increasing in proportion to the transaction load.

Latency also remained relatively stable throughout the test range until the system reached the highest tested throughput.

Predictable scaling can be particularly valuable for financial institutions planning capacity requirements. Rather than making large infrastructure investments based on uncertain future demand, banks can use more predictable performance characteristics to plan for growth.

Zero Data Loss During Failure Testing

Performance was only one part of the benchmark. Icon Solutions and MongoDB also tested how the platform responded when components failed while transactions were being processed.

Under a live load of 3,500 transactions per second, IPF recovered from an ungraceful application-node failure in less than 90 seconds. A planned node shutdown was recovered from in approximately 60 seconds. In both scenarios, the benchmark reported zero data loss and no manual intervention was required.

IPF's architecture uses persisted events and shard rebalancing to redistribute in-flight transactions across available nodes and restore transaction state.

For payment providers, this type of resilience is critical because infrastructure failures cannot simply result in lost or duplicated transactions. Financial institutions need payment systems capable of maintaining transaction integrity even when individual components become unavailable.

Cloud-Native Architecture for Instant Payments

The Icon Payments Framework was designed around a cloud-native, event-driven architecture. The company positions IPF as a payments development framework for financial institutions that require scalability, resilience, and continuous availability.

The framework is already used by major financial institutions, including Citi, NatWest, and BNP Paribas, according to Icon Solutions.

Its architecture is intended to support the changing requirements of payment environments as more transactions move from traditional net-settlement systems toward instant payment schemes.

Why Instant Payments Need Scalable Infrastructure

The continued adoption of instant payments is creating new technology requirements for banks.

Traditional payment systems were often designed around scheduled processing and settlement windows. Instant payment infrastructure operates differently, requiring institutions to process transactions continuously while maintaining availability and meeting strict service-level requirements.

Higher transaction volumes also create sharper peaks in demand. As more account-to-account payments move to instant rails, banks need infrastructure that can scale without introducing unacceptable latency or operational risk.

Icon Solutions and MongoDB argue that this environment makes resilience and scalability fundamental requirements rather than optional capabilities.

MongoDB Atlas Supports the Data Layer

MongoDB Atlas provided the database infrastructure used in the benchmark. The test involved two MongoDB Atlas clusters supporting IPF's Event Journal and Operational Data Store.

At the 6,000-payment-per-second level, these systems collectively handled approximately 430,000 database operations each second.

The results demonstrate the importance of the data layer in high-volume payment processing. A payment platform may be capable of accepting thousands of transactions per second, but its underlying database infrastructure must also maintain the performance, availability, and consistency required to support those transactions.

Building Resilient Payment Systems

The benchmark comes at a time when financial institutions are under increasing pressure to modernize payment infrastructure.

Banks must support faster transaction processing while also managing fraud controls, regulatory requirements, transaction validation, funds availability, and operational resilience. These requirements create a need for payment architectures that can handle multiple processing steps without compromising speed.

The IPF benchmark demonstrates how an event-driven payments framework and a cloud database platform can work together to address these requirements at significant transaction volumes.

What the Benchmark Means for Banks

The results provide several important takeaways for financial institutions evaluating modern payment infrastructure.

First, the benchmark showed that IPF could process 6,000 complete payment transactions per second. Second, the platform maintained a mean end-to-end latency of 0.51 seconds at that throughput. Third, the environment demonstrated near-linear scaling as transaction volumes increased. Finally, the system recovered from application-node failures while maintaining transaction integrity and recording zero data loss.

Together, these results point toward a model of payment infrastructure designed for continuous, high-volume processing rather than periodic transaction workloads.

A Stronger Foundation for the Instant Payment Era

The shift toward instant payments is changing the technical expectations placed on banks and payment providers. High throughput alone is no longer enough. Payment infrastructure must also deliver predictable performance, resilience, scalability, and transaction integrity.

The Icon Solutions and MongoDB benchmark provides a practical demonstration of these requirements. By processing 6,000 end-to-end payments per second with sub-second average latency and zero data loss during tested failures, IPF and MongoDB Atlas showed how modern payment technology can support demanding real-time workloads.

As instant payment adoption continues to grow, financial institutions will increasingly need architectures capable of operating continuously under unpredictable transaction loads. The benchmark suggests that cloud-native, event-driven platforms could play an important role in helping banks modernize their payment infrastructure while preparing for the next phase of real-time financial services.

FinTech News shares the latest trends and insights on fintech, digital banking, payments, AI in finance, and spend management.