FIA Expo: Start-Up Levyx Opts for Scalnyx to Scalable Matching Engine With Zero Data Loss

Levyx Inc., provider of high-performance, low-latency data engines optimized for SSD and hardware such as FPGAs, advanced NVM storage, and multicore processors, announced that its database engine was selected by Scalnyx, developer of AI-powered trading solutions.

Scalnyx’s ScalMatch matching engine achieves a zero data loss architecture using Levyx’s helium storage engine, which takes advantage of storage technologies like Intel Optane in advanced financial applications. The combined solution was recently featured at the FIA Expo 2019 where Levyx was honored to be selected as one of 20 start-ups to exhibit at the Innovators Pavilion.

The solution solves a industry challenge by enabling a highly-available trading platform that both: 1) minimizes infrastructure cost, and 2) meets vital performance requirements, such as latency less than 20µ-seconds, high-volume throughput of more than one million orders/sec, and stability and low jitter.

The solution is  for global market exchanges. In addition to financial trading, matching engines and the need for them to process increasingly large amounts of data are found in applications ranging from ad technology to interactive gaming to Smart Cities.

Charly Bechara, Ph.D. and CEO, Scalnyx, said: “Levyx’s focus on accelerating data processing in high-compute applications on the latest storage systems and devices made them the obvious choice for our next-gen matching engine. By leveraging their optimizations with Intel Optane we are able to add critical ‘always on’ data persistence to the solution.“

Reza Sadri, CEO, Levyx, said: “We are a big believer in the advancements that Scalnyx is making that are designed to improve the resiliency and performance of the major trading exchanges. Our technologies align well together to reduce risk and make a more robust matching engine that addresses a significant industry need.“

Matching engines are at the center of electronic exchanges and use algorithms to allocate trades among competing bids and offers. They match bids and offers to complete trades. When dealing with trading orders of clients (or members for an exchange), these systems need to include and manage all new orders received (in real time), as well as ensure that in case of failure, the systems can either rely on a continuity solution (hot backup) or restart (cold backup) with minimal loss of data. Since hardware failure may lead to loss of data held in RAM, adding resiliency and data persistence to matching engines in a way gives an exchange a major advantage over traditional architectures that rely on the redundancy principle.

ScalMatch is a performance and deterministic ULL (Ultra Low Latency) matching engine. It is built using Scalnyx C++ actor model runtime, which is battle-proven by Euronext core trading architecture Optiq. When running the matching engine on the Intel Cascade Lake multi-core CPU architecture, Helium software was used to create a performance zero data loss trading architecture. This resulted in only a 10µs increase in latency in the performance of the engine with persistence (i.e. with helium on Optane DIMMs) compared to without persistence (i.e. without helium). This differential varies by about 1µs whether the system is processing 5K orders/s or up to 800,000 orders/s.

Therefore, the Scalnyx-Levyx solution was able to add critical risk management features to the exchange’s matching engine(s) with only a negligible trade-off in performance. A full description of the use case, test parameters, and results can be found here.

In addition, a community version of helium is available.

Helium is also available for enterprise consumption by contacting the company.

Resources:
Online demo
Video