October 2023 Summaries

The benchmarking study conducted by BenchANT compares the performance, scalability, and cost-effectiveness of MongoDB Atlas and ScyllaDB Cloud, two prominent NoSQL databases, under various workloads. ScyllaDB demonstrated superior performance, outshining MongoDB in 132 of 133 measurements, with up to 20 times higher throughput and significantly lower latencies, particularly for update operations. This study used the DBaaS offerings of each vendor to ensure a fair comparison across different workloads, including social, caching, and sensor applications, with data set sizes ranging from 250GB to 10TB. ScyllaDB's distributed architecture provided nearly linear scalability and a better price-performance ratio, up to 19 times more cost-effective than MongoDB, depending on the workload and data size. The results validate ScyllaDB's efficiency for high-performance applications requiring predictable low-latency operations, although future studies are needed to assess performance with advanced data models or complex operations.

ScyllaDB Cloud has announced several updates for October 2023 to meet the growing demand for its fully-managed database-as-a-service model, offering users the ability to independently resize clusters through an enhanced Admin API, which facilitates scaling without external help. The new Terraform provider integrates ScyllaDB Cloud into larger infrastructures, and the introduction of Personal Tokens, replacing legacy API keys, enhances API security. Single-Sign-On (SSO) capabilities using SAML enable streamlined authentication across applications, while encryption at rest with Bring Your Own Key (BYOK) adds a security layer for data protection. Additionally, ScyllaDB Cloud has expanded its availability to new regions, including AWS Israel and GCP Delhi, and supports new AWS instances in Sydney and Mumbai, enhancing its global reach and infrastructure flexibility.

The podcast discussion between Lee Atchison and Dor Laor explores the decision-making process for choosing between SQL and NoSQL databases, emphasizing the importance of understanding workload requirements and anticipated usage patterns. They note that SQL databases are preferable for smaller workloads with high flexibility and query capabilities, while NoSQL databases are better suited for high-performance, large-scale applications requiring active-active setups and high availability. The conversation highlights the challenges and considerations in transitioning between different database models, such as moving from relational databases like MySQL to NoSQL options like Cassandra or ScyllaDB, and introduces the concept of NewSQL and Distributed SQL as alternatives. The discussion also examines specific performance tests comparing ScyllaDB to a Distributed SQL option, showcasing ScyllaDB's superior throughput and latency in handling massive data volumes.

P99 CONF 2023, a virtual conference focused on low-latency and high-performance distributed computing, hosted participants from around the world, offering 60 presentations distilled from over 160 submissions. The event featured a variety of technical talks, including Dor Laor's session on ScyllaDB's shard-per-core architecture and Jon Haddad’s insights on performance troubleshooting. Highlights also included Kerry Osbourne's strategies for solving complex performance issues and Mark Gritter's exploration of latency sources. The conference facilitated vibrant live discussions, notably a panel featuring Glauber Costa, Carl Lerche, and Jarred Sumner. Presentations covered topics such as Rust's advantages in backend development, innovative application instrumentation methods, and data product service level objectives. The event, attended by nearly 18,000 peers, was praised for its engaging format that allowed presenters to interact with attendees in real-time, and concluded with a keynote by Paul McKenney on understanding the Linux-Kernel Memory Model.

P99 CONF Day 2 featured a diverse lineup of presentations and discussions focused on performance optimization, open-source development, and database technologies, drawing significant engagement from its participants. Highlights included Gwen Shapira's talk on achieving high performance on a budget, Bryan Cantrill's insights into corporate open-source anti-patterns a decade later, and presentations by Stefan Johansson on reducing latencies and Tammy Everts on real-world performance budgets. The event also showcased technical explorations such as Travis Downs' session on async frameworks and Danny Kopping's case study on scaling Memcached at Grafana Labs. Attendees benefited from rich discussions and had the opportunity to engage with industry experts, with the conference fostering a strong sense of community among its 18,000 participants and generating extensive online interaction.

ScyllaDB, a high-performance NoSQL database, emerged from the founders' initial attempt to create an operating system to enhance database performance, which led them to rewrite Apache Cassandra in C++. The objective was to maintain Cassandra's robust interface while improving its efficiency and handling at scale, positioning ScyllaDB as a competitor to both Cassandra and DynamoDB. ScyllaDB offers multiple deployment models: an open-source version, a more feature-rich ScyllaDB Enterprise for self-managed environments, and ScyllaDB Cloud, a fully managed SaaS solution preferred by most new customers for its ease of use. The company's approach to solving scalability problems and its flexibility in catering to both simple and complex use cases have contributed to its growing adoption, with initiatives like ScyllaDB University and Docker-compose apps supporting new users in getting started.

ScyllaDB, a company with significant roots in Israel, expresses deep sorrow over recent tragic attacks in the country, affirming the safety of its employees and their families. Despite these events, the company remains committed to maintaining its service quality for customers, leveraging its global, predominantly remote workforce across over 20 countries. In Israel, ScyllaDB continues operations in a manner similar to its approach during the COVID pandemic, allowing flexible work arrangements. The upcoming P99 CONF event will proceed as a fully virtual gathering, while other events like DevOpsDays TLV and an internal engineering summit have been postponed to accommodate travel challenges and the needs of Israeli employees. The company encourages open communication with its representatives for any questions, conveying solidarity with all those seeking peace and performance.

P99 CONF 2023, hosted by ScyllaDB, is a free online conference that targets engineers focused on achieving low-latency performance in distributed data systems. The event features a diverse agenda that includes discussions on SQL and NoSQL databases, ORMs, infrastructure tuning, event-driven architectures, AI/ML feature stores, and benchmarking, among other topics. Highlights include sessions on deterministic distributed systems by TigerBeetle, Rust-based ingestion services at Sentry, low-latency online feature stores at Lyft, and the history of Oracle tracing. Additional sessions cover building low-latency machine learning systems, cost-effective burst scaling, and unique storage models for real-time data streaming. The conference also addresses advanced topics such as rigorous database benchmarking, optimizing ORM interactions, and unconventional methods for identifying bottlenecks in data pipelines.

"Database Performance at Scale" is a free, open-source book authored by ScyllaDB employees and contributors, including Felipe Cardeneti Mendes, Piotr Sarna, Pavel Emelyanov, and Cynthia Dunlop, which aims to provide comprehensive insights into optimizing database performance for various systems, not just ScyllaDB. The 270-page book, accessible under the Creative Commons Attribution 4.0 International License, addresses the complexities of database performance by offering practical recommendations based on extensive engineering efforts and real-world deployments. While the book is written by those affiliated with ScyllaDB, its content is relevant to users of any database system, such as MongoDB, MySQL, Postgres, Cassandra, and DynamoDB, focusing on reducing latency and throughput issues. The authors emphasize an opinionated approach, presenting their perspectives and recommendations on performance-critical decisions and optimizations, all while maintaining an engaging and accessible style. The book is available digitally for free and in print at cost, and the authors encourage discussion of its topics for specific use cases, with upcoming announcements for those interested in ScyllaDB-specific content.