June 2025 Summaries

ScyllaDB has introduced a significant architectural shift with the implementation of "tablets," representing the smallest replication unit in the database, which allows for more elastic and dynamic data distribution. This change is facilitated by the Raft consensus protocol, enabling faster scaling and topology adjustments without the need for extensive maintenance tasks like node cleanup. Tablets distribute data by splitting tables into smaller logical pieces, dynamically balancing them across the cluster, which results in cost savings and improved utilization of infrastructure. The system is designed to handle cyclical workload patterns, allowing for on-demand scaling and the ability to manage heterogeneous setups, thereby optimizing performance by distributing data based on available compute resources. ScyllaDB’s latest drivers are now tablet-aware, enhancing query routing efficiency and providing performance benefits, while future updates are expected to introduce features like transparent repairs and temperature-based balancing to further enhance cluster performance.

ScyllaDB X Cloud is introduced as a next-generation, fully-managed database-as-a-service that offers enhanced flexibility and cost-efficiency through its elastic architecture, designed to handle variable workloads with consistent low latency. Key features include rapid scaling capabilities, allowing the database to adjust from 100K to 2M operations per second within minutes while maintaining single-digit millisecond latency, and achieving up to 90% storage utilization, thereby reducing the need for overprovisioning and infrastructure costs. The introduction of tablets, which split data into smaller logical pieces, enables automatic and granular scaling based on usage patterns, while support for mixed instance sizes allows for more targeted resource allocation. Additionally, network cost optimizations such as file-based streaming and dictionary-based compression further enhance efficiency. Accompanying these technical advancements, the new "Flex Credit" pricing model provides cost-effective options by combining the flexibility of on-demand pricing with the benefits of an annual commitment, enabling users to manage costs effectively during usage spikes.

A new DynamoDB cost analyzer was developed to help developers better understand and estimate the real costs associated with their workloads, addressing the common issue of unexpected high bills due to misjudging usage, especially under on-demand pricing models. While AWS provides a pricing calculator, it often fails to capture the intricacies of real-world workloads, prompting the creation of a more nuanced tool hosted as a JavaScript single-page application on GitHub pages. This calculator not only aids those comparing DynamoDB with ScyllaDB but also offers value to anyone seeking more accurate cost assessments. It incorporates user-friendly features such as visual traffic representation, adjustable inputs for baseline and peak usage, and URL query parameters for easy reference and sharing. The tool also considers additional cost factors like network transfer and DynamoDB Accelerator (DAX), with further insights to be shared in an upcoming webinar. The overarching aim is to enable users to explore various "what-if" scenarios, helping them to better manage and predict their DynamoDB expenses.

Blitz, a startup offering personalized gaming coaching for popular titles like League of Legends, restructured its backend to improve performance, scalability, and cost-efficiency by transitioning from Postgres and Elixir to ScyllaDB and Rust. As their user base expanded, the original architecture faced challenges in handling the increasing data volume, prompting a shift to ScyllaDB for its ability to manage high write throughput and horizontal scalability, alongside a move to Rust for its memory safety and robust ecosystem. This overhaul significantly reduced infrastructure complexity and costs, replacing extensive microservices with a minimal setup that still meets performance goals, leading to a more efficient system with enhanced real-time data processing capabilities for improved user experiences.