SOURCE: Qubole

Qubole

November 22, 2017 08:30 ET

Qubole Saves AWS Customers $140M in 2017

QDS compute efficiency saves AWS users up to 80 percent in total cost of ownership

SANTA CLARA, CA--(Marketwired - Nov 22, 2017) - Qubole, the big data-as-a-service company, today announced that AWS users have seen $140 million in total cost savings with the Qubole Data Service (QDS), the industry's first autonomous data platform. Qubole has also added optimized support for AWS per second billing, meaning QDS users will see even higher compute efficiency, and announced a new online tool for AWS EMR which allows customers to estimate their potential total cost of ownership (TCO) savings.

"Qubole currently processes more than 750 petabytes of data in the cloud each month, and a significant amount of that data is run by our customers through AWS. QDS features are intelligently designed to provide the most efficient way to process big data on AWS, and the cost savings our customers have seen in 2017 show that unequivocally," said Ashish Thusoo, CEO and cofounder, Qubole. "Our AWS customers see an average reduction of 50 percent in TCO, with some customers seeing as much as an 80 percent reduction in costs."

Big data processing in the cloud takes advantage of the separation between compute and storage vs traditional on-premises big data systems that rely on converged compute and storage. Qubole has created numerous innovations that adapt open source technologies including Spark, Hadoop, Presto and Hive to take full advantage of cloud architecture and optimize the compute resources for processing data. By utilizing compute resources more efficiently, customers pay only for what they use and avoid overpaying or over-provisioning, which is commonplace today. TCO reductions are achieved through three approaches:

  • Cluster Lifecycle Management automatically starts and terminates clusters on-demand, which reduces the cost associated with idle clusters. Workload Aware Auto-scaling dynamically adds and subtracts compute nodes in running clusters by predicting workload requirements. This reduces the cost of over-provisioned clusters (provision to peak) which is common in big data deployments.

  • Spot nodes and heterogeneous clusters provide a policy-based management system and the ability to optimize price, performance and availability by mixing heterogeneous (non-identical) machines types in clusters. This reduces cost by using the highest price/performance machine types to process workloads and taking advantage of highly discounted 'spot' nodes.

  • Container packing and elastic storage reduces the number of nodes in a running cluster by compressing lightly-loaded nodes and reducing new nodes added as a result of storage overruns. This allows for more aggressive downscaling to further reduce costs.

"As we made the transition to the cloud, Qubole's ability to automate the infrastructure and easily scale to meet the demands of our users saved us time, resources and reduced our TCO by over $700k," said Wade Warren, SVP Global Engineering and Tech Ops, Wikia.

For more information, please visit Qubole's website.

About Qubole
Qubole, the leading cloud-agnostic, big-data-as-a-service provider, is passionate about making data-driven insights easily accessible to anyone. Qubole is building the industry's first autonomous data platform. The cloud-based data platform, Qubole Data Service (QDS), removes the burden of maintaining infrastructure and enables customers to focus on their data. QDS is context-aware, self-managing, and self-learning to deliver unbeatable agility, flexibility and total cost of ownership. Qubole customers process nearly an exabyte of data every month. Qubole investors include CRV, Lightspeed Venture Partners, Norwest Venture Partners and IVP. For more information visit www.qubole.com