Storage

How to Free Your Read-Intensive NoSQL Workloads from Legacy Constraints

By Doug Rollins - 12.6.18

When we introduced the Micron 5210 ION Enterprise SATA QLC SSD we narrowed the affordability gap between HDDs and flash.

The Micron 5210 SSD uses QLC to store four bits in each NAND cell (33% more than the prior generation). This really helps drive a more approachable price point for flash.

Let’s take a look at the Cassandra NoSQL performance advantages of the Micron 5210 SSD versus legacy storage (10K RPM 2.4TB hybrid HDDs) in a four-node example Cassandra cluster.

SQL 1

Unleash Your Data-Hungry Workloads

We wanted this comparison to be real-world so we designed the test process to closely emulate a real-world deployment. We also wanted to show a scalable comparison, so we used a four-node Cassandra cluster. We focused on standard YCSB benchmarks (to make the results interesting and broadly applicable).

We built each four-node test cluster and hosted a 1TB Cassandra database on it. One cluster used Micron 5210 SSDs and the other used HDDs. We set the replication factor to 3 (this means that there are three copies of the data and the cluster can sustain the loss of two data nodes while continuing operation).

How well did the Micron 5210 SSD-based cluster do? Turned out, really, really well! Here’s a quick snapshot showing relative performance (including reads, updates, inserts, read/modify/write, etc.) of each configuration across four read-centric workloads with different thread counts (taller is better).

SQL 2  SQL 5 

More Responsive Clusters, Too!

NoSQL workloads aren’t just about speed - a ton of deployments depend on consistent responses to queries (lower read latency), too.

We wanted to test responsiveness as part of our comparison, so we also compared the read latency (since we’re really talking about read-centric workloads here) at thread counts 64, 128, 192.

How did the Micron 5210 SSD-based cluster do in the responsiveness tests? We saw vastly lower response times from the Micron 5210 SSD cluster. Take a look below for these results. Since this is a response time comparison, lower is better.

SQL 3 SQL 6

 

How To Use These Results

Emerging applications and workloads are hungry for data—driving new storage demands. The Micron 5210 SSD helps meet these demands with:

  • Immediate access to vast numbers of user profiles
  • Real-time analytics and business information
  • Active archives storing immense data volumes
  • Lightning-quick VM backup and restore
  • Multinode read-heavy databases
  • Real-time media delivery
  • Content organization (tagging, writing a very small amount of data)

Read-focused workloads excel with the Micron 5210 SSD, changing the read-centric application landscape and unleashing emerging workloads to deliver results like never before. The Micron 5210 SSD enables capacities measured in multiple terabytes (TB), read throughput in hundreds of megabytes per second (MB/s) and density-driven cost effectiveness for high-capacity, ultra-fast, read-focused storage to enable new design opportunities and performance thresholds.

Many emerging workloads share a common thread: they read far more data than they create, and they must read it quickly and efficiently. When these workloads are cost-sensitive, we’ve had to compromise, relegating them to slow legacy storage, which compromises results.

The Micron 5210 SSD helps bridge the affordability gap, ushering more read-centric workloads into SSDs to drive them further.

Learn more about the Micron 5210 SSD, QLC technology, and the transformative effects on your business today:

Micron 5210 SSD: www.micron.com/5210 

QLC Technology: https://www.micron.com/products/advanced-solutions/qlc-nand 

Stay up to date on what’s trending in storage by reading Micron’s Storage Blog and following us on Twitter @MicronStorage and connect with us on Linkedin.

Doug Rollins

Doug Rollins

Doug is a Senior Technical Marketing Engineer for Micron's Storage Business Unit, with a focus on enterprise solid state drives.
+