Here at ETERBASE, we believe that the sign of a good exchange is one that can deliver just as seamless an experience during periods of increased trading activity as it does during times when the market is relatively calm.
We’ve seen throughout the past few years that many exchanges are found wanting when the market sees a spike in activity levels, resulting in severe slow-downs and outages.
In our quest to provide our customers with a smooth trading experience regardless of market activity levels the design team at ETERBASE identified four primary factors that would drive the technical and architectural choices;
The more that a single node in the processing chain can process within a single second, the fewer nodes required to carry out the requested overall throughput.
However, it’s worth noting that the overall throughput of a series of processing steps is only as fast as its slowest node, so the performance of the system is reflected directly in the cost and quality of the entire solution.
Every piece of equipment in the process must be able to deliver to a very high standard, which is precisely what the ETERBASE trading platform provides for its customers:
- An order matching component developed in-house that can deliver up to 15 million trades per second on a single node.
- A messaging solution that can deliver up to 20 million messages over the network between two nodes.
Although many exchanges often claim to employ algorithmic and high-frequency trading practices, for successful application these methods require very low latency.
A London-based multilateral trading facility renowned for its delivery of sub-millisecond latencies with their state of the art custom-built components, LMAX have been delivering on this front since 2011.
Since then, however, we’ve seen several (some even open source) components become available that not only closes the gap on the technology used back then but also surpass it in many ways.
This makes it much easier to offer similar sub-millisecond latency performances, and shorten the time-to-market.
In comparison to some of the exchanges built only a few years ago, the difference in latency is remarkable.
Our analysis has demonstrated that some of the more popular messaging systems can introduce latencies as high as 50-11 milliseconds with each hop, with performance worsening even further during times of higher usage.
To keep latency low, ETERBASE have decided to;
- Develop in-house critical processing chain components with a particular focus on latency
- Utilize a simple, fast messaging solution that can deliver latencies under 100 microseconds, even during times of high use
- Carry out most of the processing in memory using a sophisticated persistence strategy
- Minimize the number of required hops as much as possible, which means not having to sacrifice scalability
Ability to partition and fault tolerance
Both of these objectives work hand-in-hand.
The greater the ability to fan out and horizontally scale up (or scale back, if required) the number of nodes for any given service, the better the exchange’s reaction to peak usage periods.
Also, should we see a fault with a node, our expertly-designed fault strategy will ensure service is not interrupted by merely replacing the damaged node with a healthy one.
As you can see, those who decide to trade on the ETERBASE exchange are going to experience a fast, seamless experience regardless of market activity levels.