SingleStoreDB Self-Managed 4: Market and Strategy CEO Q&A

SingleStoreDB Self-Managed 4 represents a leap forward in relational database innovation with new capabilities such as real-time geospatial intelligence and high-throughput connectivity to Apache Spark. It also includes a free forever, unlimited capacity Community Edition, enhancements to the Optimizer for distributed joins, and a new version of the SingleStore Ops management framework.

Start Using SingleStore Community Edition Now
Unlimited scale and capacity. Free forever.
Download Community Edition →

As Big Data abounds, understanding every company and product can be tricky. To make easier, here are a few questions and answers about SingleStore.

The database landscape is big. Where does SingleStore fit?

SingleStore is the leading database for real-time transactions and analytics. That means we’re operational by nature, much more akin to SQL Server, Oracle, or SAP HANA than Hadoop. SingleStore is also:

• In-Memory
  • Providing the utmost performance for today’s demanding workloads
• Distributed
  • Enabling cost-effective, horizontal scale-out on-premises or in the cloud
• Relational and multi-model
  • Allowing companies to use in-house SQL tools, applications, and knowledge
  • With JSON and Geospatial data formats supported
• Software
  • Designed to run on commodity hardware for costs savings

You talk about transactions and analytics. Can you explain in more detail?

To meet real-time demands, companies must be able to capture information across millions to hundreds of millions of sensors or mobile applications. They also want to analyze that data up to the last transaction. So with real-time operations you don’t have the luxury or the pain of ETL. You need to bypass ETL by transacting and analyzing in a single database designed to support these concurrent workloads. It boils down to analytics on changing datasets, today’s critical capability.

SingleStoreDB Self-Managed 4 includes the SingleStore Spark Connector and the SingleStore Loader for HDFS and S3. How should we think about SingleStore with Spark and Hadoop?

SingleStore and Spark work well together as they both have memory-optimized, distributed frameworks. Spark is a processing framework that enables real-time transformation and advanced analytics, but Spark itself does not have a storage or persistence ability. By storing data permanently in SingleStore, customers get an easy way to build operational applications, and the ability to take operational data and ‘round-trip’ it to Spark for advanced analytics.

With Hadoop, customers frequently build simplified Lambda architectures using SingleStore. All data can go directly to HDFS for long term archiving. Simultaneously, data can go directly into SingleStore, bypassing Hadoop for the real-time path. Should historical data be needed for analysis, SingleStore can import that data from HDFS using the SingleStore Loader.

Many folks say not everything needs to be in-memory. How do you respond?

We agree! While in-memory computing remains critical for many applications, the pace of data growth still eclipses memory-only solutions. This is exactly why SingleStore ships with an integrated column-store optimized for disk and flash. Now customers can create tables entirely in-memory, or across a combination of memory and disk and flash. And starting with SingleStoreDB Self-Managed 4, we license the software based on DRAM capacity so the use of disk or flash storage in the column store is unlimited at no additional cost.

We are big believers that customers will see the benefit of placing data in a structured format from the beginning, and the SingleStore column store will let them do that affordably.

You have always focused on SQL. All the while NoSQL has received a lot of discussion. Where is SingleStore in this?

SQL is the lingua franca of the data processing world, having been well adopted over its decades-long history. Companies can build SQL applications quickly, and then immediately use in house analytics tools and practices to derive insights.

Prevailing wisdom used to be that SQL could not scale. That is untrue and companies are now discovering the powerful combination SingleStore delivers of a relational, in-memory, distributed database with speed, scale, and simplicity.

The future is multi-model and SingleStore also includes JSON and Geospatial datatypes to fulfill this promise.

The Community Edition is freely available but not open-source. What strategy is SingleStore pursuing?

Our strategy is to build a scalable software business. This means running on all the platforms customers want to use, including public and private clouds, in the way they want to use them. Community Edition is a way to allow more people to use SingleStore without limitations on time or capacity.

There is a wide spectrum of commercial database offerings. On one extreme there are proprietary hardware / software combinations. Our belief is that proprietary hardware is a down elevator. On the other end of the spectrum there are open-source projects with businesses built around them. But that is essentially selling consulting hours, not software. There are many successful businesses in the middle of the spectrum, such as MySQL’s dual commercial / open-source licensing, and we maintain open-source projects, including SingleStore Loader for Hadoop and S3, a large number of general-purpose Python packages, and the SingleStore Spark Connector.

At SingleStore, we want to offer performance that trumps legacy vendors, deliver it a fraction of the cost, and provide complete deployment choice across hardware, data centers, and clouds. In doing so we hope to help more companies achieve their goals to become real-time enterprises.