Optimizing Apache Cassandra™ Database at Scale Solution Brief

The Solution for Cassandra at Scale

HGST has a long history of producing the largest and most reliable hard disk drives (HDDs) in the world. Couple those hard drives with the broad SanDisk-brand portfolio of flash storage technologies and you get a solution that enables Cassandra architects to achieve optimal Cassandra clusters. While modern multi-core servers allow parallel execution of multiple threads, Cassandra was not originally written to fully exploit them, which often leaves cores idling, waiting for data from storage. Ultimately, the low utilization causes server sprawl and wasteful spending of IT budgets.

Apache Cassandra™ database at scale can use both the cost-effective capacity of HGST-brand Ultrastar® helium hard drives and the density and performance capabilities of SanDisk®-brand solidstate drives (SSDs) to fully exploit flash and modern servers and to provide optimal performance and consolidation.

What SanDisk- and HGST-brand drives can do for Cassandra:

• Store more data in the same or smaller footprint
• Minimize query and data operation times on critical datasets
• Customer-optimized price and performance
• Reduce or eliminate JBODs and controllers

About Apache Cassandra

Cassandra is an open source NoSQL database written in Java and specifically optimized to be scalable, decentralized, fault tolerant, and, above all, performant. It is used at some of the web’s largest properties and throughout financial services and other industries as a repository of record with multiple petabytes of data under its control.

As a NoSQL database, Cassandra was built from the ground up for scale-out architectures. Instead of investing in a large, centralized database server with massive amounts of storage and memory capacity, architects can deploy more modestly configured servers to perform the same types of operations and guarantee the same levels of uptime and data reliability.

Scale-out provides a powerful method for increasing database performance and capacity. Need more compute power? Add servers to distribute the workload. Need additional storage capacity? Add servers and rebalance. Yet all of these server additions, if not properly managed and minimized, can lead to a classic case of server sprawl with massive operational expenses from large and underutilized server farms.

Summary

Cassandra can be a powerful tool to store and extract value from massive amounts of data. However, like any scale-out tool, it needs to be applied carefully and thoughtfully, or it can result in a massive server sprawl and associated headaches.

For the largest Cassandra databases, adding HGST Ultrastar helium HDDs in industry-standard, fully serviceable chassis is ideal. This solution provides high capacity and good performance in a small footprint, and it enables the construction of cost-effective, massive clusters.

Ideal candidates for SanDisk SSDs are Cassandra databases in which queries take too long to return data or in which applications are not meeting their SLAs. In these cases, SanDisk SSDs, potentially in a directto-CPU connected NVMe interface form factor, may dramatically reduce query response times and allow you to maintain or reduce your server footprint at massively increased query performance.