Adaptive designs, builds, and operates front-office trading technology for financial services and capital and commodity markets. Although Adaptive was cloud native from the beginning and has been an AWS Partner since 2014, many of its early solutions catered to clients operating with on-premises data centers. To meet clients’ growing demand for cloud-native trading systems, Adaptive began to use Aeron, an open-source messaging system that can be deployed on cloud-native and on-premises infrastructures.
The core of high-performance trading systems is the messaging component, which facilitates the rapid and secure exchange of trade orders, confirmations, and other vital data. The messaging system’s speed and reliability can make or break a trade, especially when transactions occur in microseconds. Today, most low-latency trading systems run on premises; however, this infrastructure can present challenges such as system outages, limited scalability, high maintenance, and connectivity costs. Aeron Open Source provided the foundation that Adaptive needed to overcome these challenges, which helped the company move trading to the cloud.In 2023, Adaptive brought to market Aeron Premium: a version of Aeron that provides financial firms with additional software components, enhanced features, and enterprise-grade support. To verify that Aeron Open Source and Premium could meet stringent performance benchmarks, Adaptive, alongside the AWS team, assessed the solutions’ capabilities in a cloud-native environment. These tests took place in the AWS Stockholm Region using Amazon EC2 C5 Instances, which deliver cost-effective high performance at a low price per compute ratio. Specifically, Adaptive used Amazon EC2 C5n Instances, which can deliver 100 Gbps of network throughput to a single instance.
Adaptive aimed to prove that running Aeron in the cloud could help capital market firms build fast trading systems that reliably process large data volumes. “Aeron provides a set of capabilities that can be layered on top of AWS to provide capital market-grade performance in throughput, latency, and, most importantly, resilience,” says Matt Barrett, CEO of Adaptive. “By harnessing the power of AWS, we’ve demonstrated exceptional latency and notably increased throughput, delivering tangible results for our capital market clients.”
Adaptive first tested Aeron Transport, the open-source version’s low-latency message transport system, to test raw network throughput and latency. Then, Aeron Cluster—the technology’s clustered storage component—was tested to demonstrate the throughput and latency that a highly available system using Aeron can achieve.
Aeron Open Source uses Berkeley sockets. Aeron Premium uses Data Plane Development Kit, whose kernel bypass helps applications directly access network and hardware resources, reducing the overhead associated with traditional kernel-based networking. This configuration dramatically improves messaging latency and increases throughput.
During the performance tests, Adaptive saw impressive, predictable latency for Aeron Transport. The open-source version achieved a round-trip latency of 66 microseconds at 100,000 messages per second for a standard 288-byte message size. Open-source Aeron Transport also yielded an impressive total throughput of 350,000 messages per second.
With Aeron Premium, latency dropped to 43 microseconds at 100,000 messages per second, 500 times faster than commonly used messaging protocols. Encrypted transport, also a feature of Aeron Premium, had a latency of 46 microseconds, which is almost 60 times faster than common encryption protocols. Throughput for the Premium version increased by 8.6 times, achieving 3 million messages per second at a latency of less than 1 millisecond in the ninety-ninth percentile.
Adaptive also tested Aeron Cluster, which provides resilience and redundancy for critical systems. A cluster leader node handles all active workloads, while follower nodes reliably replicate the leader and seamlessly continue operations in the event of a failure or planned downtime of the leader node. For Aeron Open Source, Adaptive achieved a latency of 235 microseconds at 100,000 messages per second. The company also saw a throughput of 250,000 messages per second.
Performance further improved with Aeron Premium: latency for fault-tolerant replication dropped by 45 percent, from 235 microseconds to 130 microseconds. Adaptive also achieved a throughput of 2 million messages per second in 181 microseconds, an eight-time improvement.
