January 2025 Summaries

Tripadvisor leverages ScyllaDB and AWS to deliver real-time personalized recommendations to its vast user base, enhancing user experience by quickly adapting to individual preferences. Using machine learning models, Tripadvisor processes over two billion requests daily, serving 400 million unique monthly visitors with tailored suggestions for hotels, attractions, and experiences. The company's personalization efforts are supported by a robust architecture that includes independently scalable microservices, a custom feature store, and a dual-database system that separates real-time data processing from offline data warehousing to ensure performance efficiency. ScyllaDB, chosen for its low-latency capabilities, replaced Cassandra due to its enhanced throughput and reduced operational demands, facilitating rapid data retrieval and processing. This infrastructure supports Tripadvisor's goal of offering seamless, relevant content to travelers, helping them plan ideal trips by utilizing advanced data engineering and machine learning practices.

A blog post by Attila Tóth details the construction of a high-performance shopping cart application using ScyllaDB, with a focus on leveraging ScyllaDB's Change Data Capture (CDC) feature to monitor and export table changes. This application uses Python and FastAPI for the backend, providing a user-friendly experience for managing products, including adding, removing, and updating product information. ScyllaDB, a NoSQL database known for its low latency and capability to handle large data volumes, is integral to the app's performance, ensuring smooth user interactions that are crucial for e-commerce success. The post emphasizes a "query-first" approach to data modeling, ensuring optimized database schemas for specific use cases and maintaining single-digit millisecond latency. It also highlights the utility of CDC in analyzing user behavior, such as tracking product additions to carts, which can be used to personalize user experiences and improve conversion rates. Additionally, CDC data can be exported to platforms like Kafka for further analytics, providing insights into user behavior and product popularity.

The Monster SCALE Summit 2025 is a highly technical, virtual conference focused on addressing engineering challenges related to performance-sensitive, data-intensive applications at a massive scale. Following an enthusiastic community response, the event will feature over 60 sessions across two days, covering topics such as multi-tenant database scaling, distributed systems, cost optimization in cloud environments, and large-scale content understanding. Notable sessions include discussions on re-engineering Postgres for multi-tenant applications, optimizing ScyllaDB for increased throughput, and exploring the evolving architecture of Atlassian's Confluence Cloud. The conference aims to provide insights and practical strategies for engineers and architects dealing with the complexities of scaling systems to manage extreme data volumes and demands effectively.

Supercell, a Finland-based gaming company known for hits like Clash of Clans and Brawl Stars, faced the challenge of developing a real-time social platform for hundreds of millions of players with just two engineers. They evolved their basic account system, Supercell ID, into a comprehensive cross-game social network infrastructure capable of managing account management, friend requests, chat, and player presence tracking across five major games. This transformation was facilitated by implementing a two-way communication system and using a hierarchical key-value store with Change Data Capture. ScyllaDB Cloud was chosen to handle the required low-latency, high-throughput event persistence. The system architecture is designed with APIs, proxies, and event routing/storage servers, ensuring operational simplicity and high performance. Supercell ID abstracts complex functionalities into a single system, allowing the development of new games to be accelerated by packaging essential features into their infrastructure.

ScyllaDB University has been actively enhancing its educational offerings, including the introduction of new lessons and updates to existing materials, particularly focusing on data modeling and application development within the NoSQL framework. The platform offers self-paced online courses and hosts live training events to improve skills in ScyllaDB, with the upcoming ScyllaDB University LIVE event scheduled for January 29th, featuring both Essentials and Advanced tracks. Recent course updates emphasize key principles such as query-driven design, efficient data distribution, and the strategic selection of primary and clustering keys to optimize performance and scalability. Lessons cover challenges like hot partitions and tombstones, and introduce advanced topics such as collections and drivers, alongside practical tools for diagnosis and prevention of database issues. Additionally, the platform provides insights into ScyllaDB's operations, including read/write paths and tracing, to ensure data consistency and high availability.