HPC when being used for High Frequency Trading is focused on maximizing performance, increasing operational efficiency and accelerating innovation through the deployment of high-speed processors, networks, applications and low latency messaging platforms. From automating the data center to leveraging cloud storage models, key CTOs and Heads of IT are exploring the latest technologies and solutions powering high-performance trading.
A few key factors driving the trading industry forward include:
Data savviness essential for survival – harvesting intelligence from data—and converting that intelligence into productive financial decisions—is the KEY driver for firms today. Still it seems that only a few are fully optimizing the conversion of raw data into actionable analytics to satisfy competitive and regulatory forces.
Quantum computing, as we know, WILL be a revolutionizing advancement when the time comes. But the question is when? Everyone I’ve spoken with about quantum says we are at least a decade away from any meaningful applications. Firms like D Wave and Rigetti Computing are leading the charge.
Parallel computing employs several processors to execute or process an application or computation simultaneously, essentially allowing for large computations by dividing the workload between more than one processor, all of which work through the computation at the same time. This principle is employed by most supercomputers.
Machine learning is an area in which many people are excited about and reaps a lot of attention, especially in the way of:
- Neural networks – a class of machine learning algorithms used for time-series forecasting, trading, risk modeling, etc. Use either supervised, unsupervised or reinforcement learning strategy. Large neural networks comprise millions, if not billions of computations.
- Unsupervised learning is computational costly and time consuming – in many cases several weeks are required to complete a training session; Unfortunately, large delays in training highly limit their usage in practice as many parameters often needs to be tested, and each test requires a full session of training.
- Big data productivity tools & software utilities – compilers, debuggers, performance analysis and visualization tools
- Deep Learning – New advances in HPC (Intel Xeon Phi, i.e.) is allowing for decreased training time from weeks to days, or days to hours, for deep learning algorithms
The recent progress in HPC for the financial vertical will continue to drive new vendors towards the capital markets and banking fields, which will only serve to make the industry even more efficient and further competitive.
Written by Steven Reichard
June 19, 2018