Database tech developer MemSQL today announced it signed a debt facility that provides up to $50 million of new capital. Co-CEO Raj Verma says it will chiefly be used to deliver new and existing products and services and to “accelerate growth” in the months to come.
AI and machine learning models require fast databases like MemSQL’s in order to perform at their peak. Organizations that lack the right technical components in their production pipelines run the risk of failure — according to IDC, 25% of brands already using machine learning report a 50% failure rate. MemSQL ostensibly prevents this with a platform that serves as the backend for fraud detection, portfolio risk tracking, and even facial recognition apps in industries ranging from financial services, energy, and government and public sector to retail and ecommerce.
MemSQL — which can be deployed on-premises, as-a-service, or a hybrid of both — works like most relational databases, which is to say it accepts requests (e.g., for a user, image, video, document, or internet of things event) in the form of queries for data contained within the database. It processes these queries and returns the results in milliseconds, after which it assigns them a score that indicates their overall quality.
MemSQL centralizes data with built-in workflows while performing queries to identify new models. Streaming ingest eliminates the need for data integration tools through built-in batch and real-time pipelines, while the compiling of queries into low-level machine code speeds up responses.
MemSQL can discover anomalies or predict events by combining real-time and historical sources, delivering instant matching to models against data sets. It applies built-in models to maximize response time while scoring models as data is ingested, and performs ad hoc analysis with business intelligence tools like Tableau, Looker, Microstrategy, and more.
MemSQL can ingest millions of events per day while simultaneously analyzing billions of rows, with support for geospatial data like area, distance, and location analytics and JSON multi-attribute objects. Data can be stored across clusters of machines with transactions written directly to disk or to memory, and with compression that optimizes resources for storing up to petabytes of data.
On the redundancy and management front, MemSQL holds a backup copy of data to protect against loss and ensure consistency, and it eliminates duplicate records at the time of ingestion. It also automates common tasks like starting, stopping, restoring, and backing up clusters and provides a monitoring interface to diagnose and assist with query, pipeline, and storage performance tuning by collecting query profiles and exposing potential bottlenecks. Users can manage security configurations by role and group and audit all activities to external secure locations, or manage existing account access to enable security tasks like tracking access.
MemSQL also offers Helios, a fully managed cloud database hosted on public clouds like Amazon Web Services and Google Cloud Platform. It starts at $2.75 per hour per unit (equal to 8 processors, 64GB of memory, and 1TB of storage), with options to support up to infinite units.
IDC expects the worldwide big data analytics market will be worth $274.3 billion by 2022, and MemSQL is considered among the pack leaders. It saw 70% growth in annual recurring revenue and single-digit cash burn last year, and its customers include Verizon, Intel, Uber, Comcast, Sony, Pandora, and Samsung, among others.
Hercules Capital served as the underwriter for the financing, which brings MemSQL’s total raised to approximately $158 million following a $30 million series D in March 2018. The nine-year-old company is headquartered in San Francisco and has offices in Portland, Oregon; Seattle, Washington; Sunnyvale, California; London, U.K.; Lisbon, Portugal; and Kyiv, Ukraine.