Financial trading applications notoriously demand the low-latency, highest performance platforms available in the marketplace and all for a simple reason – for traders, time is money. In other words, improve the rate at which banks and trading firms can ingest, analyze and access financial data and this will generate greater alpha – or higher profits for them. In no other industry (that we can think of) can better technology be used to drive costs down (better bottom-line) and impact top-line revenues. Interestingly, despite being the domain of “real-time” necessities, trading operations are often still relying on antiquated and expensive systems and software in their back-end infrastructure. Queue the entry of Levyx’s data engines, optimized for Intel® Optane™ technology, that are powering leading edge financial applications, ranging from Distributed Ledger (DLT) based FX data solutions, to cloud-based securities processing solutions, to order management caching, to time series databases, to options based risk analytics and backtesting solutions. In general, these all enable traders to run their businesses more efficiently and drive higher profits.
How frustrating is it that your financial institution will spend millions improving the hardware components of your trading platform every year but have these investments only have a marginal impact on how your trading improves year over year? Faster processors, more processors, more memory, larger nodes, more nodes – how effective are any of these approaches to helping you make more money? As an intelligent storage engine, Levyx’s software brings out a leading performance of your hardware and let’s you take full advantage of better storage devices like Intel Optane technology. Specifically, it allows the application programs to treat space on Intel Optane DC Persistent Memory as a viable replacement for RAM in popularly used platforms like Python Pandas and Apache Spark (i.e. Amazon EMR, Cloudera, IBM, etc…) as well as traditional C and Java based trading applications. This results in two major benefits: first, better hardware utilization and, second, greater performance due to the bare-metal performance. Performance gets a lift due to Levyx’s compiler and engines which optimize for data locality and bypass the operating system kernel for many operations allowing data to be fed with very low latency into applications including into hardware accelerators like GPU and FPGA. No more letting your trading operations get bogged down by slow system software.
Levyx’s software is designed to be especially good at managing, retrieving, processing, and indexing of large datasets (think 100GB to Petabytes) It bypasses the file-system and kernel buffers and directly performs SSD-friendly I/O to reduce I/O latency and maximize I/O bandwidth for storage devices. It uses just-in-time compilation to generate analytics code on the fly, executing that code on bare metal for maximum efficiency, low latency, and throughput. The generated code performs direct I/O via Intel Optane DC Persistent Memory to reduce read/write amplification.
Optionally, it can push complex calculations to hardware accelerators like Intel’s advanced picture archiving and communication system (PACs). Levyx’s technology uses the latest ISA, lock-free data structures, and multicore design principles to maximize system resource utilization—for example, reading input data from Intel Optane Solid State Drives (SSDs) and creating a stream of input data into a group of PAC accelerator cards.
All of the use cases referenced above share the same fundamental need to access very quickly a large Hot tier of data to meet the business needs. The combination of Intel Optane technology and Levyx is unique in the industry, the result of 6 years of intensive R&D led by a team of PhD’s that are solely focused on these solutions.