London-based Adaptive Financial Consulting acquired low-latency trading tech software company Real Logic earlier this year, and is now gearing up to market Real Logic’s low-latency, open-source messaging technology, Aeron.

Adaptive CEO Matt Barrett says adoption of Aeron by capital markets firms has been slow because Real Logic never marketed or promoted the tech at all.

“They were a group of individuals being paid to implement features within Aeron on behalf of clients, and putting that work into the open-source world. They were doing that on an ad hoc basis or via support contracts and had no interest in growing,” Barrett says, adding that with the acquisition, which was completed in February, he hopes Adaptive can leverage its marketing and brand to grow usage and make Aeron more widely applicable and adopted.

Adaptive, using its technical accelerator Hydra Platform, builds trading systems like single-dealer platforms, matching engines, and execution venues for clients in financial services, including exchanges, banks, and investment managers. The consultancy had a long-running relationship with Real Logic before the acquisition and chose Aeron to underpin Hydra’s architecture when it began developing the platform about five years ago.

Aeron itself was built in 2014 by veterans of the low-latency messaging space Martin Thompson and Todd Montgomery, for a large US exchange that wanted an open-source messaging system. Thompson was a co-founder and CTO of FX (and now crypto) exchange LMAX, where he had helped develop an influential open-source architecture that could handle lots of applications at low latency. Montgomery was CTO at 29West, which was acquired by Informatica in 2010, where he built a commercial messaging solution, Ultra Messaging. With the Adaptive acquisition, Thompson and Montgomery joined the Adaptive management team.

“We’ve had a long-term relationship with Real Logic; we knew them personally through various jobs going back 10 or 11 years. And when we started investing in Hydra about five years ago, we had a commercial relationship with them to provide support for the underlying open-source pieces,” Barrett says. “So our clients that were using Aeron were fully covered with support all the way down. But as Hydra grew, we felt that the combined proposition of our offering plus Aeron should be sitting in the same organization to provide fully aligned support to clients.”

Aeron is not one thing, but rather an overall brand name for a couple of technologies. One aspect of Aeron is Transport, a brokerless, User Datagram Protocol (UDP)-based messaging layer for unicast or multicast. It is comparable to commercial solutions like Informatica’s Ultra and Tibco FTL and open-source solutions like RabbitMQ and ActiveMQ. Unlike brokered messaging or middleware over TCP—like the open-source Kafka, for example—many low-latency messaging systems used in financial services tend to be over the UDP because it is faster.

But it’s Aeron’s Cluster that Barrett believes is the most innovative aspect of the tech stack. Cluster essentially sequences client messages, replicates them, and sends them to multiple back-up servers. Each server gets the same sequence of messages, so the same state can be replicated across all the machines in a group.

Cluster is based on the Raft consensus algorithm. Consensus algorithms help keep replication consistent among a group of distributed machines, allowing them to operate coherently as a group, and continue to operate even if some of the group members fail. These algorithms play a key role in software for enterprise-scale systems that must be fault-tolerant—that is, large systems that must continue to function even if individual components fail.

Cluster, then, is useful for building for stateful applications such as one might find in the exchange world—central limit order books, matching engines, smart order routers, protocols like request-for-quote (RFQ)—where continuity of service is paramount even when there is a hardware or network failure.

“A client could have its matching system running on Aeron Cluster. And if one of the nodes, one of the machines that Cluster is running on, fails, the entire system keeps running,” Barrett says.

Cloudy communications

Barrett says this kind of tech is important in an era when financial services firms are accelerating their cloud migrations. “In the old days, infrastructure deployed on-premises, like into the datacenter of a bank, was reliable—it lasted for a long time and it didn’t fail very frequently. But now as we move to the cloud, there are a lot more failure points, lots of different places where systems could go wrong, and you need to have far more resilient application architectures that can handle that sort of thing.”

Multicast is the financial sector’s method of choice for streaming market data, as it scales easily and is meant for getting the same data to a select and often large group of clients. However, the public cloud does not support multicast delivery. This means that financial firms must build their own internal networks to handle it, and exchanges still run their core infrastructure—matching engines and feeds—on their own server farms or premises.

Amazon Web Services (AWS) does have a multicast service in the cloud for its Transit Gateway. But none of the cloud providers offer anything as performant as what firms achieve with on-premises multicast, and so cloud multicast is not suitable for anyone whose business relies on ultra-low latency.

Aeron, however, can run on the cloud and enable low-latency messaging between co-located facilities, as well as on-premises and in deployed infrastructure, Barrett says.

“Multicast was, when you had physical infrastructure, a very sensible way to go. But now as you move to the cloud, it is far less performant, if available at all,” Barrett says. “In conversations we have with cloud service providers, they go back and forth on how they think about fully supporting multicast; no one is suggesting there will be support for users to take multicast-dependent messaging infrastructures they have deployed, and simply move them to the cloud and have everything work. That’s not going to happen.”

As Aeron is open-source, Adaptive doesn’t license the software, but rather monetizes its association with the brand, providing support, training, and consulting around it. Adaptive also sells a proprietary set of extensions that run on top of Aeron to make it faster and more secure, tools for high-throughput messaging on the cloud, and encrypting messaging traffic.

Open-sourcing technology has its benefits, of course, and probably most important of these is that a technology can attract a solid, expert community that works on maintaining and improving the tech all the time. Aeron being open-source does have one drawback for Barrett, however, particularly as the vendor is focusing efforts on promoting the tech: It’s difficult to tell what kind of market penetration the tech has achieved.

“The problem with Aeron being open source is that anyone can pick it up and start using it. So we hear about that more anecdotally than we do formally,” Barrett says. “Every now and then, users will get in touch and ask for help with running it in production, and our response is, ‘Oh wow, that’s amazing, what are you guys doing with it?’

Adaptive is not the only vendor with a long association with Aeron. TransFicc, a connectivity provider for fixed income and derivatives, also promotes its use of Aeron. And as an open source technology, it seems to be an active and growing community: Its Github shows that it now has 96 contributors and almost 760 forks.

WatersTechnology asked a few technologists in the low-latency trading space for their opinion of Aeron, and they were all positive about the tech itself. One of these was Luc Burgun, CEO of field programmable gateway array (FPGA) tech vendor NovaSparks, who said in an email: “Regarding Aeron, our CTO, Pierre Gardrat, is quite enthusiastic about it. Actually, we developed a prototype for distributing market data from an FPGA with a similar reliable protocol based on UDP in the past.”

However, others point to the lack of community around Aeron. Adaptive competitor Scott Logic wrote a blog post in 2020 (long before the acquisition) asking the question, “Is Aeron a good choice for a messaging solution?” The post’s answer was yes, but noted its lack of community as a major drawback.

“The most overt problem with using Aeron is simply how little is known of it. Despite being around for about five years, there is a lack of online content surrounding how to use it,” the blog notes, adding that a Google search of Aeron is more likely to lead one to an “iconic” brand of high-backed office chair.

Adaptive will soon be hosting events at its offices in London and New York, where current and potential clients can share use cases, as well as other engagements that vendors usually do when they have a product to promote, such as attending conferences. Barrett will be hoping that increased interest in open-source financial tech, and Cluster’s approach to fault tolerance in an era where resilience is more important than ever, will let the technology speak for itself.