May 2018 Summaries

We raised $30 million in Series D funding led by GV and Glynn Capital, bringing total funding to $110 million. SingleStore is driving digital transformation for major companies like Comcast, Pandora, and Uber due to its cloud-architected database that prioritizes high-speed analytics and scalability. The new funds will be used to expand teams and continue record-breaking growth. SingleStore aims to revolutionize relational databases by enabling data consumption as a utility, allowing customers to derive more value from their data and reduce infrastructure costs.

SingleStore demonstrates high throughput performance for ad-hoc analytical queries, processing 1,280,625,752,550 rows per second on a cluster of 448 Intel Skylake cores. The query's success is attributed to careful code design, exploiting distributed execution, instruction-level and data-level parallelism. To achieve this performance, SingleStore uses pipelining and SIMD (Single Instruction, Multiple Data) techniques. Pipelining allows the processor to execute multiple instructions in parallel, while SIMD enables the processing of multiple rows simultaneously. The query's success also depends on the efficient organization of data, which is stored in a columnar format with a dictionary-based indexing system. This approach enables grouping by small integers and reduces write-write conflicts. Overall, SingleStore's performance is made possible by leveraging parallelism through distributed execution, pipelining, and SIMD, allowing for a significant increase in computing power to respond to queries within a short time frame.

SingleStore is a foundational datastore that incorporates machine learning functions in three ways: calculations outside the database, calculations on ingest, and calculations within the database. SingleStore offers an open source Spark Connector for integrating ML and AI outside the database, enabling high-throughput and bi-directional operations with Spark and TensorFlow. This integration allows for real-time machine learning scoring and predictive analytics, such as powering wind turbines' sensor information to predict equipment failures. Within the database, SingleStore enables machine learning through extensibility, including stored procedures and algebraic operations like DOT_PRODUCT, which can compare vectors for similarity or execute elements of a machine learning pipeline quickly. This allows organizations like Thorn to improve their real-time image recognition capabilities and build a machine learning recommendation engine in SQL.