Announcing Firebolt Engines: Next-Generation Compute Infrastructure for Cloud Data Warehousing

Firebolt Inc., a cloud data warehouse provider, announced its next-generation compute infrastructure, Engines. Designed to address the pain points of today’s traditional data warehouse platforms, Firebolt Engines redefine how data is ingested, processed, and queried, offering high performance, low latency, and cost efficiency.

Customers use cloud data warehouses (CDW) to aggregate data from multiple sources and perform large-scale analytics on this data. There has been an exponential increase of this data in the past decade due to various factors across different verticals. Existing CDW solutions have addressed this rapid increase in scale with techniques such as decoupling compute from storage and leveraging object storage, which can store potentially unlimited amounts of data. But, with advancements in deep learning and generative AI, modern analytic workloads are becoming increasingly AI-driven, seeking actionable insights in near-real time and need granular access to large-scale data. The infrastructure provided by existing CDW solutions either does not meet the needs of these low-latency workloads or costs so high that the price-performance of the workloads outweigh the benefits. These workloads are also becoming  more dynamic, with unpredictable demands on their infrastructure, which must be flexible enough to rapidly scale based on these demands. For such dynamic workloads, traditional CDW solutions provide the ability to either scale up (to a bigger size) or scale out (add more nodes). But, both these scaling mechanisms have limitations - while scale-up is constrained by the maximum hardware resources that can be made available on a single node, certain types of scale-out do not work well for concurrent queries even with additional nodes. In addition, with existing solutions, customers are forced to double their resources every time they scale out their compute resources, increasing their costs and negatively impacting the price-performance of their workloads. For many customers running business-critical workloads, any service downtime for planning and managing software upgrades adds to their operational burden. For some of these customers, such service downtimes can impact their Service Level Agreements (SLA), further increasing their costs resulting from the SLA violations. These modern analytic workloads need a next-generation infrastructure that can not only handle data at scale,  but also provide low-latency and high performance in a cost-effective manner.

Firebolt’s state-of-the-art infrastructure, Engines, provides key capabilities that meet the demands of today’s large-scale workloads. An Engine comprises one or more clusters, and each cluster is a collection of nodes of a certain type. Engines provide multi-dimensional elasticity, with the ability to dynamically scale along any of the three dimensions: 1:/ Scale Up by using a node type that offers more hardware resources  2/ Scale Out by adding more nodes per cluster and 3/ Concurrent Scaling by adding more clusters to an engine. With Concurrent Scaling, an engine can distribute more queries across multiple clusters, enabling customers to efficiently deal with a sudden increase in the number of queries or number of users. The ability to scale across all dimensions, including concurrency scaling, is available to all users and not restricted to a specific edition. Firebolt provides granular scaling, which allows customers to incrementally scale their engines one node at a time. While they can easily double the number of nodes, Firebolt customers have more flexibility to add only the number of nodes their workload requires, significantly reducing their costs and improving the price-performance of their workloads. Note that customers can change both the number of nodes and the node type with a single scaling operation. Software upgrades in Firebolt are automatic and incur zero downtime for the customers. The periodic updates provided by Firebolt, including security updates, bug fixes and performance enhancements are delivered seamlessly and transparently, minimizing the operational overhead for the customers and ensuring their workloads are not interrupted. 

Engines add a huge amount of needed flexibility in the way we provision query power without causing a bottleneck or costing us development cycles. Our application has low latency requirements but with a highly variable number of active users throughout the day. The multi-dimensional scaling provided by Firebolt Engines will allow us to easily deal with this variable number of users, keeping the user experience smooth while minimizing the costs for both compute and development time. Win win!

- Eden Trainor, Tech Lead of Machine Learning Engineering at Compass Digital Labs
‍

For more information, visit firebolt.io/elasticity