InBrief Philip Howard – Research Director, Information Management www.VoltDB.com 209 Burlington Rd #203, Bedford, MA 01730, USA Tel: (978) 528 4660 VoltDB Email: [email protected]

The company CREATIVITY SCALE VoltDB was founded in the first decade of this century by Michael Stonebraker and others. The eponymous database the company offers was based originally on H-Store, a research project that emerged in 2007 and which was available as an open source project. By 2012 VoltDB had come to market with a commercial version of the product and the H-Store closed down “ in 2016. VoltDB is the logical VoltDB has its headquarters in

choice for a cloud- Bedford, Massachusetts and also EXECUTION TECHNOLOGY deployable, transactional has offices in the UK and China. It database that can flexibly has a direct sales force but also The image in this Mutable Quadrant is derived from 13 high level handle high-volume data leverages partners that OEM its metrics, the more the image covers a section the better. Execution metrics relate to the company, Technology to the streams for service providers technology. While the product is product, Creativity to both technical and business innovation and to monitor and leverage Scale covers the potential business and market impact. in real time. suitable for a wide range of use Openet ” cases, roughly three quarters of the company’s customers are within How does it work? the telecommunications sector (Nokia, VoltDB uses a shared-nothing architecture to Vodaphone and Deutsche Telecom are examples), achieve database parallelism, with both data with financial services (Barclays) its second largest and processing distributed across all the CPU user base. There is a Community Edition of the cores within the servers composing a VoltDB product, which is free to download, as well as an cluster. In other words, the product should scale Enterprise Edition. The product is also available via as the number of cores increases, not just by AWS. the number of servers. Horizontal partitioning also enables scalability. SQL access is supported What is it? via pre-compiled stored procedures. The VoltDB is a NewSQL database. That is to say, it looks unit of transaction is the stored procedure, and feels like a relational database, and it provides which is Java interspersed with SQL. Further, the ACID compliance and immediate consistency stored procedures are executed in the that you would expect from a relational database, containing the relevant data, which means that but it has a different underlying architecture it is possible to eliminate round trip messaging compared to a traditional relational product, which between SQL statements, which is another reason enables horizontal scalability, high availability and for the performance of VoltDB. extreme performance. More generally, VoltDB leverages in-memory On a very simplistic level, VoltDB removes techniques to improve performance, but also all of the baggage that has accreted in relational supports spill-to-disk capabilities where memory database systems over the last forty years. In capacity is insufficient. For data that is merely particular, it does not have latching or locking, and warm or cold, as opposed to hot, VoltDB provides it doesn’t need buffer management because it uses exporters, as can be seen in Figure 1. As can be modern in-memory processing techniques. These seen in this diagram, VoltDB supports ingestion processes alone take up significant amounts (we from streaming sources (Apache Kafka, Amazon have heard estimates of over 90%) of processing Kinesis plus various message buses) and export time in traditional OLTP databases before you even to data lakes and other dedicated analytic start to do anything else. environments. From the latter, machine learning

© Bloor 2019 Architecture Flexibility Concurrent analytics In-process analytics Criticality Performance Ease of use Scale

models may be imported into VoltDB Why should you care? using PMML (predictive modelling There are multiple ways to support hybrid mark-up language), where they are We conducted“ environments that support both transaction treated as stored procedures. While a global search and processing and analytics with low latency in both we applaud the support for PMML selected VoltDB because cases. Some of these approaches will be more as an interoperable standard, it we believe it’s the world’s familiar than others and we would argue that fastest and smartest is unfortunately true that many familiarity and the ability to use SQL are major in-memory database.” machine learning and AI models are benefits. Conversely, you need a distributed Mahindra Comviva not PMML compatible, so we would architecture to support environments such as the like to see VoltDB supporting models Internet of Things, where a substantial amount of built by language (R, Python, Scala and so processing is needed at the edge. VoltDB has the forth). advantage of meeting both of these requirements, While on the subject of machine learning with ACID compliance, low latency and – likely – and stored procedures, it is also worth noting lower cost of ownership. that some types of analytics have fairly static requirements compared to machine learning, The Bottom Line which is typically dynamic. An example, would For all intents and purposes VoltDB should be compliance with MiFID II. We comment upon be thought of as a (distributed) relational this because one of VoltDB’s customers has database. Making the move to VoltDB, whether embedded the Drools rules engine into VoltDB as as a replacement for an existing system, or for a stored procedure precisely as a mechanism to deployment within a greenfield environment, will enforce compliance with MiFID II. be a non-disruptive choice that is worth serious Finally, VoltDB supports containers and consideration. Kubernetes is used for orchestration. Multi-site active/active deployments are available.

Figure 1 – VoltDB exporters

FOR FURTHER INFORMATION AND RESEARCH CLICK HERE

© Bloor 2019