High-performance computing is all about maximizing throughput, killing latency, and getting the most out of your hardware. This session focuses on FinTech and AdTech use cases such as security and advertisement exchanges, low-latency tactics and related ethics/privacy.
Todd Montgomery is a networking hacker who has researched, designed, and built numerous protocols, messaging-oriented middleware systems, and real-time data systems, done research for NASA, contributed to the IETF and IEEE, and co-founded two startups. He currently works as an independent consultant on high performance systems and is active in several open source projects, including Agrona, Aeron, ReactiveSocket, and the FIX Simple Binary Encoding (SBE).
10:35am - 11:25am
If we are lucky, we think of performance as an afterthought. Same with other “non-functional” requirements such as security, quality, scaling, etc. However, experience shows that it is easier, at least to a small degree, to think of these “anything but non-functional" requirements up front. This often leads to better designs, better systems, and simply cleaner solutions. This session will explore the everyday things that those with an eye to performance and efficiency do that can be leveraged by anyone to build better software faster.
11:50am - 12:40pm
Today’s fintech applications require resiliency, fault-tolerance, and extreme performance. In an area traditionally dominated by proprietary software and bare-metal computing, open-source technologies and cloud computing present exciting possibilities for the future.
Deploying to the cloud offers many benefits such as multi-region redundancy and elastic capacity, but comes with less control of the underlying resources. Can advances in cloud networking technology, coupled with cutting-edge OSS, provide the kind of latencies demanded by modern applications? What are the factors that need to be considered when trying to squeeze every last drop of performance from these deployments?
In this talk, we will explore the improvements in cloud networking technology, look at performance testing and measurement in cloud environments, and outline techniques for low-latency messaging from an application and operating-system perspective. Finally, we will compare the performance of the latest cloud tech with a bare-metal installation.
1:40pm - 2:30pm
At Intent our machine learning platform processes real-time and historical data to predict user intent on billions of page views a month. At the heart of this system is a serverless data pipeline that allows us to gather, process, store, and analyze data from disparate data sources. In this talk, we’ll discuss the motivations of switching to a serverless infrastructure, and lessons learned while building and operating such a system at scale; focusing on operability, stability, scalability, and ease of development.
2:55pm - 3:45pm
Typical large Adtech platforms handle 99th percentile response latency under 10ms and throughput in millions of requests per second while generating TBs of logs per day. Detection of invalid traffic (fabricated by bots) is fraught with many of the same engineering challenges to satisfy similar low-latency and high-throughput requirements. While an Adtech platform aspires to maximize ad relevancy, the goal of fraud detection is to minimize such invalid traffic. By design, both types of platforms are high-performance distributed systems that must cooperate to be successful.
In this talk, we will explore a set of core building blocks exhibited by Adtech platforms and apply them towards building a fraud detection platform. Our claim is by leveraging Cloud and OSS we can drastically reduce the time to production and infrastructure cost. Thus, bare-metal and proprietary software should no longer be deemed a barrier of entry.
After addressing performance, we will touch on the key attributes of system reliability and quality in an Adtech system. We will conclude with many insights learned from building one of the leading fraud detection platforms from the ground up.
4:10pm - 5:00pm
As more and more applications are distributed, latency is becoming a critical factor in delivering a good user experience. Techniques used in Fintech and Adtech – such as zero copy, flyweights, composite buffers, pooled memory, shared memory transport, direct memory in languages like Java – can be applied to help improve performance. Many of these techniques were applied while developing the open source RSocket protocol. The session will include examples and a demonstration of how these techniques can be used to improve performance in distributed applications.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.