July 2023 Summaries

Jaeger, a popular open-source distributed tracing system developed by Uber Technologies, is widely used for monitoring and troubleshooting microservices-based applications. The integration of ScyllaDB, a high-performance NoSQL database, as a storage backend for Jaeger enhances its performance and scalability, particularly in high-traffic environments. This setup enables Jaeger to handle a larger number of spans per second, optimizing the performance and resource efficiency of distributed applications. The blog post details the implementation process using Docker and Git, highlighting a proof of concept where ScyllaDB outperformed Apache Cassandra by collecting approximately 120,000 spans within 60 seconds. This demonstrates ScyllaDB's superior scalability and resource utilization, making it an ideal choice for supporting Jaeger's distributed tracing capabilities. The post concludes by emphasizing the benefits of using ScyllaDB to achieve significant performance improvements for Jaeger's monitoring functions, offering valuable insights into the execution path of requests and enhancing the user experience in complex, large-scale systems.

The ScyllaDB Community Forum serves as a hub for both new and experienced users of ScyllaDB, the high-performance NoSQL database, to exchange knowledge and troubleshoot issues, with recent discussions covering a broad range of topics. Participants have explored the feasibility and impact of maintaining up to a trillion records in a single table, the practical limits on the number of tables in a keyspace, and the differences in CPU consumption under various Linux kernels after executing specific commands. Other discussions have focused on the redistribution of data when nodes are added or removed, potential performance degradation with large row sizes, and the evolution of Cassandra’s data model from schema-less to schema-based. Additionally, participants have debated strategies for querying tables efficiently, such as using paging to minimize latency.

The blog post by Cynthia Dunlop explores a range of cutting-edge NoSQL database features and practices designed to support systems where low latency at scale is essential. It highlights the experiences of various companies, such as Numberly's transition to Rust for improved performance in their Kafka and ScyllaDB-based data pipelines, ShareChat's database modernization to handle rapidly increasing user engagement efficiently, and Optimizely's strategies for maximizing database concurrency during high-demand periods. The text also delves into the significance of robust data modeling, the challenges of media streaming systems, and the considerations involved in adopting Database-as-a-Service (DBaaS) solutions. Through real-world examples and expert insights, the post provides valuable lessons on optimizing NoSQL databases for performance, scalability, and cost-effectiveness.

The blog post discusses the importance of selecting appropriate storage solutions for ScyllaDB deployments, emphasizing that storage is a critical component affecting database performance. It highlights common issues users face, such as performance problems due to suboptimal disk choices, and the need for understanding how databases interact with storage to avoid deployment failures. The post explains how ScyllaDB uses an I/O scheduler and Seastar’s Scheduling Groups to prioritize tasks and manage system resources effectively. It details the evolution of ScyllaDB's I/O scheduler and its current capabilities, which include sophisticated self-throttling and cross-shard capacity sharing. The piece also covers storage-related best practices, including the use of locally attached NVMe SSDs and RAID-0 setups for optimal performance and explains the importance of regular disk evaluations and the use of tools like scylla_io_setup and scylla_fstrim. The author dispels misconceptions about locally attached disks, highlighting their cost-effectiveness and reliability within a replicated database environment.

ScyllaDB is transitioning from a Vnode-based replication strategy to a new algorithm centered around tablets to enhance data distribution and elasticity across clusters. While Vnodes have traditionally been used to evenly distribute data by randomly selecting token space split points, they often result in imbalances and inefficiencies, particularly for small tables. The tablet-based approach allows for a more dynamic and flexible distribution by assigning fixed-size tablets to serve table fragments, enabling the movement of individual tablets to balance loads. This new method also integrates the Raft consensus algorithm, offering advantages like strong consistency and higher throughput, although it introduces challenges such as potential latency during leader failover. Tablets facilitate more efficient data management, offering easier resharding and cleanup processes, and improve load distribution by enabling finer control over where data resides. As ScyllaDB implements this technology, it aims to address the limitations of Vnode partitioning by providing a more scalable and manageable system.

ScyllaDB University LIVE is a free, virtual event scheduled for July 25, 2023, offering both beginners and advanced users an opportunity to deepen their knowledge of ScyllaDB, a high-performance NoSQL database. The half-day event features live, instructor-led training sessions divided into two parallel tracks: ScyllaDB Essentials and Advanced Topics and Integrations. Participants can expect to learn about ScyllaDB architecture, data modeling, and cluster management, as well as advanced strategies for optimizing deployments and migrating to ScyllaDB Cloud. The event includes interactive sessions, an expert panel for Q&A, and opportunities for participants to complete quizzes, hands-on labs, and earn completion certificates. Attendees are encouraged to complete the ScyllaDB Essentials course beforehand to maximize their experience and understanding of the ScyllaDB technology.