March 2018 Summaries

** In this recent webcast, SingleStore shared tips and tricks for operationalizing SingleStore, covering topics such as permissioning a new cluster, configuring memory limits, monitoring health metrics, taking backups, adding nodes securely to the cluster, and more. Attendees were able to ask questions and receive answers on various aspects of SingleStore, including backup file accessibility during restore, deploying a new Master Aggregator, hardware failure mitigation, node addition in High Availability mode, and secure practices for clearing memory buffers. The webcast also covered best practices for columnstore use cases and provided information on how to access the recording and get more support from the SingleStore team.

Query optimization is a process used by databases and data warehouses to rework user queries to deliver quick responses with minimal compute resources. It involves algorithms that sift through multiple query plan choices to make the best decision, often relying on cardinality estimates, costing, and understanding data distribution. Query optimization aligns with AI principles, using predictions to identify efficient ways to run operations. The process benefits enterprises by providing faster results, reducing compute resources, and supporting more users. As distributed systems become more prevalent, query optimization becomes even more important, considering network usage and join algorithm choices. In the future, AI techniques will continue to play a significant role in query optimization, enabling real-time adjustments and improving database performance over time.

SingleStore processing has been demonstrated to achieve a trillion rows per second on the latest Intel Skylake servers, providing interactive response times under a quarter of a second for users. This is made possible by SingleStore's technology for executing single-table group-by/aggregate queries on columnstore data incredibly fast, using methods such as operations directly on encoded data, compilation of queries to machine code, vectorized execution, and use of Intel AVX2 SIMD enhancements. A cluster with 448 total cores was used to achieve this record-breaking performance, and the results show that SingleStore can deliver real-time analytics on large data sets without precalculating results, enabling more flexible interaction from users and encouraging them to explore the data.

SingleStore has integrated Full-Text Search (FTS) into its product, a highly requested feature that allows customers to build real-time applications with improved performance and relevance. FTS uses an inverted index data structure, stemming capabilities, and relevance scoring to provide faster search results compared to regular database queries. The integration with SingleStore is straightforward, leveraging the open-source Lucene technology, allowing users to specify columns for full-text indexing when creating a table. With FTS, users can mix structured and unstructured query types in their SQL queries, making it easy to take advantage of the feature without requiring extensive knowledge of full-text searching. The management of the full-text capability is transparent to the user, with automatic sharding, failover, rebalance, node addition and removal, backup/restore, etc., handled by the system. Users can utilize FTS for various use cases, such as autocomplete functionality, mixing structured and unstructured query types, and retrieving relevance scores in their queries.