STAC Summit, 1 Nov 2017, NYC

Vendors with exhibit tables at the conference. (click here for the exhibitor list).

• Peter Lankford, Founder & Director, STAC

Peter will present the latest benchmark results for big-compute workloads based on derivatives risk.

• Peter Lankford, Founder & Director, STAC

Peter will discuss the latest benchmark results involving big data workloads such as tick analytics and backtesting.

• Dave Rukshin, CTO, WorldQuant
• Rene-Paul Lafarie, Customer Engineer, Finance, Google Cloud Platform
• Tom Golway, Chief Technologist, Finance, HPE
• David Cohen, Senior Principal Technologist, Intel

Beefing up analytics is top of the agenda for many capital markets firms today. Whether it's searching for alpha or better execution, supporting new risk-management needs, improving surveillance, or any number of other challenges, demand is soaring for both data-intensive and compute-intensive analytics. Which software and hardware architectures are capable of providing the scale, speed, and flexibility that today's quants and machine-learning algorithms require? Can firms meet those demands at a reasonable cost? How is the role of on-prem and cloud infrastructure evolving? We'll put these questions and more to a cross-functional panel of experts.

• Thomas Rodgers, Member of ISO C++ Committee's SG1 and SG14 study groups

As the capabilities of processors and co-processors continue to increase, the opportunity for applications to take advantage of those capabilities gets larger--but so does the programming challenge. As part of research at the Illinois Institute of Technology into parallel software design patterns, Tom has been working with the latest Intel Xeon Phi processor (Knights Landing) to understand how best to exploit its features. Using QuantLib as a use case, he is adapting the code to leverage AVX-512 instruction sets, Phi-specific intrinsics and mixed thread/vector concurrency code generation in the Intel Compiler Collection, as well as High-Bandwidth Memory, streaming writes, and Sub-NUMA Clustering. In this talk, Tom will provide a snapshot of his experience so far.

• Ehsan Totoni, Research Scientist, Intel Labs

Firms using Python for financial modeling want to scale their analytics with maximum efficiency and minimum effort. To address these challenges, researchers at Intel Labs developed a compiler-based system called High Performance Analytics Toolkit (HPAT) and contributed it to the open source community. HPAT compiles data analytics and machine learning code written in Python (Pandas/Numpy) to efficient parallel binaries with MPI, while requiring minimal code changes. According to Ehsan, the lead contributor to HPAT, performance can be orders of magnitude faster than Apache Spark in cluster/cloud environments. Ehsan will explain how HPAT works and what Python programs it supports.

• Thomas Rodgers, Member of ISO C++ Committee's SG1 and SG14 study groups

Getting the lowest latency or highest throughput from C++ code remains a top priority for many trading firms. That explains the enthusiastic response to our talk by Bjarne Stroustrup last year and our panels on C++ in HFT at the Chicago and London STAC Summits. One of those panelists was Tom Rodgers, who has over 20 years of experience writing high-performance C++ code for trading firms and is on the ISO C++ Committee, including SG 1 (concurrency) and SG14 (low latency). After providing a quick review of the structure of modern machines, Tom will discuss fundamentals such as the cache, alignment of data, and data placement. He will cover the importance of structuring code to give the compiler the most information possible for optimization, and he’ll show the differences in machine code that result from optimal versus non-optimal code. Finally, he’ll touch on key performance issues in multi-threaded design, challenging some common assumptions along the way (e.g., is lock free always better?).

• Peter Lankford, Founder & Director, STAC

Peter will provide his perspective on the preparedness of firms for MiFID2 time compliance, as well as how regualtory enforcement is looking at this point. He will also provide updates on STAC-TS research on time synchronization, timestamping, and event capture, as well as software tools to demonstrate compliance with time-sync regulations.

• Stephane Tyc, Co-founder, McKay Brothers

We’re now a good five years into the low-latency wireless era, which followed just two years after the advent of the ultra-direct fiber era. Is there another “era” to come? Does some other transport promise to move latencies asymptotically closer to the speed of light in a vacuum, while raising the table stakes even higher for low-latency trading? Or could a new approach potentially reduce the cost of ultra-low latency enough to broaden competition in the elite speed game? What about satellites? Drones? Balloons? In this talk, Stephane will review some of these approaches and analyze their potential to influence the trading business.

• Peter Lankford, Founder & Director, STAC

Peter will discuss the latest research and Council activities related to low-latency/high-throughput realtime workloads.

• Miguel Rivera, Former CTO, Sun Trading LLC
• Matthew Knight, VP of Technical Marketing, Metamako [Slides and video]
• Olivier Baetz, COO, NovaSparks [Slides and video]
• Laurent de Barry, Founder & Managing Director, Enyx [Slides and video]

As the latency race continues to compress the time delta between successful and unsuccessful trades, FPGA-based solutions continue to play an important role in the architecture of many trading firms. In this discussion involving panelists with different angles on FPGA, we'll get perspectives on a number of questions, such as: What is the state of the art for FPGA hardware, and what trajectory can we see for future products? Do the hardware suppliers share the same product priorities as trading firms, or is there a widening gap? What is changing and what is staying the same in how firms are deploying FPGA-based solutions? How are leading firms handling the productivity challenge of FPGA DevOps? Come hear our panelists' answers and ask some questions of your own.