October 2021 Summaries

In the last year, SingleStore has made significant progress in its rebranding efforts, expanding its product offerings to all public clouds and marketplaces, and increasing its customer base by 60%. The company has also grown its employee base from 200 to nearly 300 people and expanded internationally into EMEA and Latin America. SingleStore has secured major funding rounds, earned industry-leading awards, and achieved notable certifications such as SOC 2 Type 2. Looking ahead, the company remains committed to providing fast and reliable database solutions for data-intensive applications and is poised to lead the way in a new class of modern databases. With a strong focus on diversity and inclusion, SingleStore aims to create a more inclusive workplace and support STEM educational programs to build a pipeline of female engineers.

In-memory databases (IMDBs) offer improved performance compared to traditional databases by storing data in RAM, reducing the need for disk access and enabling faster query processing. IMDBs are particularly useful for applications that require rapid response times, such as game leaderboards or real-time analytics. However, they can struggle with large amounts of data, concurrency, and analytic workloads. SingleStore is a modern database that leverages an in-memory architecture while also supporting data warehouse workloads through its tiered storage model, providing fast performance for both transactional and analytical applications.

The SingleStore team refactored their Visual Explain feature to improve its user experience, simplify maintenance, and reduce code complexity. They chose React Flow as the new library for visualizing query plans, which provided a better node graph layout, customizable rendering, and performance improvements. The refactor aimed to address issues with the previous implementation, such as a non-fluid movement, limited zoom levels, and unnecessary logic for zooming and panning. The team worked with the Product Design and Query Optimization teams to gather feedback and ensure the new feature met user expectations. Despite some challenges, the refactored Visual Explain achieved a better user experience while reducing maintenance efforts.

Many organizations have invested in specialized database technologies to support big transactional and analytical workloads. General-purpose database technology was previously sufficient, but new data requirements demanded more specialized capabilities. Organizations now face challenges with multiple databases, including unbridled copying of data, integration issues, high latency, and the need for combined analytics. To address these issues, unified databases, also known as translytical or HTAP (Hybrid Transaction/Analytical Processing) databases, offer a solution that can handle both workloads concurrently, support multiple data models, and combine different types of analytics in a single query, leading to less copying of data, simpler architectures, access to zero-latency data, and combined analytics.

Distributed SQL databases offer a consistent developer experience, improved performance, and scalability compared to traditional approaches such as sharding middleware and NoSQL. They provide the benefits of ANSI SQL, ease of application connectivity, and ACID-compliant transactions without compromising on data consistency. Unlike shared-everything database architectures like MySQL and PostgreSQL, distributed SQL architecture is made up of three layers: a SQL API, distributed query execution, and distributed storage. The primary characteristic of a distributed SQL database is the ability to shard automatically based on a declared key, reducing the overhead of sharding middleware previously discussed. Distributed databases are horizontally scalable and elastic, with managed service offerings enabling one-click scaling, offering high availability and data resilience without complexity. Cloud-native distributed SQL databases also provide unlimited storage and replication by default, ensuring data persistence in object storage.

The SingleStore 7.5 release introduces limitless point-in-time recovery (PITR), a mission-critical capability for enterprise customers, allowing them to recover a complete database to any point in time over a pre-defined window. This feature benefits database administrators, executives, and developers by eliminating the need for manual backups, reducing the risk of data loss, and enabling continuous online backup. It simplifies recovery from database corruption or other problems, makes it easy to use production data in development, and allows users to function as a system of record with near-zero recovery point objective (RPO) and recovery time objective (RTO). The feature is particularly useful for banking customers who need to maintain seven years of historical data. With limitless PITR, developers can test their applications on up-to-the-minute production data, reducing the need for manual backups and improving development efficiency.