March 2015 Summaries

The company SingleStore is showcasing a new demonstration of real-time geospatial location intelligence, titled Supercar, at the Gartner Business Intelligence and Analytics Summit in Las Vegas. This demonstration uses a dataset containing 170 million real-world taxi rides to simulate the ability to monitor and derive insights across hundreds of thousands of objects in real-time. By integrating geospatial datatypes into its relational database, SingleStore enables simple queries to generate informative results, such as calculating passenger counts or trip distances. The demonstration combines SQL queries with a developer-focused mapping platform from Mapbox, allowing users to pan across the map and zoom in on specific sections to run queries. This showcases SingleStore's capabilities in providing native geospatial intelligence for real-time analytics and decision-making.

We are thrilled to be in Las Vegas this week for the Gartner Business Analytics and Intelligence Summit, where we will showcase our new geospatial capabilities and demonstrate how Pinterest is using SingleStore and Spark for real-time analytics. We have a variety of activities planned, including games and giveaways, happy hour with SingleStore CEO Eric Frenkiel, and sessions on topics such as the future of data warehousing and choosing the right DBMS for digital businesses. The event also features a featured session from Eric Frenkiel, where he will discuss how moving to real-time pipelines means replacing batch processes with online datasets that can be modified and queried concurrently. We invite attendees to stop by our booth #119 and schedule a meeting or demonstration with SingleStore while at the show.

In recent years, there was a surge in interest for NoSQL solutions due to the challenges of scaling traditional relational databases. However, as reality set in and workloads became more complex, many companies began to incorporate SQL and relational models into their data management strategies, finding that they provided the necessary structure and performance. The relational model has kept pace with modern workloads, particularly thanks to advancements in in-memory computing and distributed systems, which have made it economically viable to store high-value data in memory and scale foundational infrastructure. Today, companies can store data in a relational format from the outset, providing the operational model and required performance with built-in structure, allowing them to focus on data science rather than data plumbing.

Scaling databases can make even simple tasks, such as counting, seem difficult. However, advances like in-memory storage have made it possible to handle high-throughput tasks at scale without the need for specialized, single-purpose databases. Counters are records that store the count of events by type and time bucket, allowing for fast querying and aggregation. A counter database can be designed using a simple schema, with tables for counters and event types, and can be populated with data using an upsert query. The database can then be queried to retrieve counts of events by type and time bucket, with various SQL queries providing different levels of summary and analysis. In testing, the database demonstrated impressive performance, achieving throughputs of 1.6 million increments per second in a simulated workload, making it suitable for high-throughput applications.

We are excited to be at Spark Summit East in New York City, showcasing our new geospatial capabilities and collaborating with Esri to demonstrate the power of SingleStore's geospatial features in conjunction with Apache Spark. Our geospatial capabilities enable enterprises to achieve greater database efficiency by integrating geospatial functions into a single database that is in-memory, linearly scalable, and supports relational SQL and geospatial functions. We launched an official news release last week, which highlights the benefits of our integration, including lock-free capabilities and powerful manipulation functions for geospatial data. At Spark Summit East, you can catch a talk by Mansour Raad from Esri on how Esri-Spark-SingleStore is used to showcase real-time geospatial intelligence, and visit our booth where we will be giving away t-shirts and Estes flying drones!

Esri showcased upcoming features for the next release of SingleStore at their Developers Summit in Palm Springs, using a large real-world geospatial dataset, such as the "Taxistats" NYC taxi ride compilation. Esri is the dominant player in the GIS market, and SingleStore aims to make geospatial data a first-class citizen, comparable to other types of data. The demonstration used heatmaps to visualize aggregate geospatial data and showed how simple queries can be performed on the data, leveraging features like GEOGRAPHY_INTERSECTS. This technology has real-world use cases for cities planners and taxi businesses, enabling them to optimize traffic flow and improve efficiency. SingleStore combines geospatial data with relational SQL capabilities in a single database that is in-memory, linearly scalable, and supports a full range of functions.

Ankur Goyal and Anders Papitto from SingleStore will present a talk titled "A Distributed In-Memory SQL Database" at the AMPLab at Berkeley on March 11th, discussing major architectural design decisions and technical details behind their product's features such as lockfree, code generation, durability/replication, distributed query execution, and clustering. The presentation aims to cover new directions for the product, including ideas on leveraging Spark. Ankur Goyal is the Director of Engineering at SingleStore and has expertise in distributed systems, compilers, and operating systems, while Anders Papitto is an engineer with experience in distributed query execution, column store storage, and other components. The AMPLab focuses on integrating algorithms, machines, and people to tackle Big Data challenges through machine learning, cloud computing, and crowdsourcing.