Paving the Yellowbrick road to closer integration with cloud data stores

Traditionally, data warehouses meant enterprises had to either commit to one platform or suffer the complexity of managing multiple quasi-compatible infrastructures. With that in mind, Yellowbrick Data recently launched Yellowbrick Manager to make it easier for administrators to manage data warehouses across distributed cloud and on-premises deployments with a simplified interface.

Users can control Yellowbrick data warehouses in public clouds, on Yellowbrick hardware instances, in private clouds, and even at the network edge, the company said. Yellowbrick Manager provides a unified control system that uses the Kubernetes container orchestration system to enable users to manage and control both cloud and on-premises deployments with enhanced performance capabilities.

“It’s this single unified control plane, together with the adoption of Kubernetes in our database software, that sets us apart,” Yellowbrick CTO Mark Cusack told VentureBeat.

The company also added agile data movement capabilities to help customers integrate Yellowbrick with data lakes built on cloud object stores like Amazon S3. A technology preview is expected in early May, and general availability is set for the second half of 2021. Also in early May, there will be an update to Yellowbrick data warehouse release 5, with data lake integration enhancements that include native support for cloud object storage (including Amazon S3 and Azure Data Lake Storage Gen 2).

Working across clouds

One of the criticisms of hybrid clouds has been around the disjointed set of technologies and user experiences across private and public clouds. There are different identity and access management approaches, which makes it harder to govern access to data across these environments, and they are typically provisioned differently. Yellowbrick is positioning itself as the first data warehouse for the distributed cloud by introducing a unified control plane that works across the most common cloud platforms. This control plane increases the simplicity of running multiple data warehouses in different physical and line-of-business locations across an enterprise.

The Kubernetes cloud native architecture provides Yellowbrick Manager with a single unified control panel to provision new data warehouse instances in different clouds, manage existing infrastructure, and monitor deployments.

Yellowbrick said Yellowbrick Data Warehouse queries run 3 times faster on Andromeda-optimized instances for private clouds than on the company’s first-generation architecture. The performance improvements are also the result of the company switching to new AMD 64-core CPUs, increasing bandwidth between server notes, and adding dual proprietary Kalidah scan accelerator cards that offload workloads such as filtering, compression/decompression, and row/column transposition from CPUs.

“The Andromeda instance is the fastest platform to run Yellowbrick on,” Cusack said.

The company is optimizing deployments to take advantage of differences across the big cloud providers. For example, if a public cloud instance has high-performance storage, Yellowbrick can adapt to the underlying hardware to take advantage of those benefits.

Yellowbrick’s approach

Yellowbrick is focusing on improving the management aspects for distributed clouds rather than joining them together directly.

Competitors tend to fall into two major categories: legacy data warehouses such as Teradata, Oracle, and SQL Server or cloud-only data warehouses like Snowflake, Amazon Redshift, and Microsoft Azure Synapse. Yellowbrick seeks to differentiate itself by taking a unified approach to addressing distributed data challenges like data gravity and sovereignty requirements. It was designed from the ground up to optimize price/performance across bare metal and virtualized infrastructure in public clouds.

Yellowbrick product marketing VP Justin Kestelyn says Yellowbrick has an advantage over legacy vendors that have been doubling down on older architectures and have a harder time with real-time analytics. Traditional cloud vendors have not been aggressively pursuing hybrid options for analytics at the edge.

“We’re winning business from all of these vendors by providing the best price/performance and the lowest deployment risk,” Cusack said.

Source: https://venturebeat.com/2021/04/14/paving-the-yellowbrick-road-to-closer-integration-with-cloud-data-stores/