葉家郡
Jia-Jun Yeh, a seasoned principal software engineer specializing in FinTech, blends deep computer science knowledge with expertise in C++ and Go. He has evolved into a skilled Kubernetes administrator and application developer, demonstrating exceptional adaptability and swift learning. Over the past few years, Jia-Jun has confidently embraced a significant leadership role, steering his team towards innovation and collective achievement with decisive direction.
Sessions
This proposal explores the innovative application of the open-source Read-Copy-Update (RCU) mechanism to optimize high-frequency trading systems. RCU is a powerful synchronization technique that enables rapid data access, which is crucial in trading environments where milliseconds can make the difference between success and failure.
We will introduce the different flavors of Userspace RCU and delve into the specific implementation details of the RCU variant we used in a real-world case study. By sharing this practical example, we will demonstrate techniques for integrating Userspace RCU to significantly improve real-time data handling and system responsiveness in high-frequency trading applications. Attendees will learn how to select the most appropriate RCU flavor and effectively incorporate it to address the unique challenges of these low-latency systems.
The overarching goal is to showcase how leveraging RCU can boost the performance of high-frequency trading, while also encouraging innovative approaches to tackle the technical obstacles inherent in this domain.
This presentation highlights a backend engineer's journey in optimizing financial data analysis by advancing from basic CRUD operations to sophisticated SQL features. Faced with the challenge of slow API responses due to extensive database interactions, a novel approach was adopted—transferring computational processes directly into the database. This shift not only streamlined data retrieval through single SQL queries but also enhanced performance by exploiting SQL's advanced functionalities like Materialized Views, CTEs, and Window Functions. The initiative not only significantly reduced API response times but also minimized development and maintenance overhead, demonstrating the power of advanced SQL in backend optimization and the importance of embracing database capabilities for system efficiency.
SIMD是一項神秘的程式設計領域，大多數人或許知道如何開啟參數來讓編譯器使用SIMD指令集最佳化，卻很少有人擁有實際的SIMD程式設計經驗。在這次的分享中，將透過開源專案的實際原始碼，來一層一層解析那些生澀的內容，學習如何利用SIMD指令集撰寫程式及加速。
I will start by addressing some of the challenges encountered during the installation process of Kubernetes, with a particular emphasis on the use of BGP Mode in CNIs. Following this, I will introduce fundamental concepts such as routing, overlay networks, BGP, and ECMP, to explain how to optimize the efficiency and reliability of Kubernetes networking.
The talk will highlight the advantages brought by BGP Mode, including improved network efficiency and performance, as well as the challenges it presents, such as the need for in-depth network knowledge and management of routing tables. To conclude, the importance of understanding these network concepts for fully leveraging Kubernetes will be emphasized, encouraging a deep dive into documentation to enhance network efficiency.