Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

Capacity Planning: In-memory Vs. SSDs for Predictable Real-Time Database Performance

DZone's Guide to

Capacity Planning: In-memory Vs. SSDs for Predictable Real-Time Database Performance

· Database Zone ·
Free Resource

Download the Altoros NoSQL Performance Benchmark 2018. Compare top NoSQL solutions – Couchbase Server v5.5, MongoDB v3.6, and DataStax Enterprise v6 (Cassandra).

By Young Paik

Real-time Web and mobile interactions are putting new demands on databases to deliver predictable performance and availability. And in many cases, success can depend on the ability to sort through terabytes of data in a handful of milliseconds to deliver the right response.

To meet the incredibly high demands these applications require, many developers are looking toward RAM-based, fault-tolerant databases. However, at Aerospike, we have demonstrated in real-world deployments that a database solution for these applications can be handled not only in RAM, but also by using much more cost-effective solid-state drives (SSDs) and still meet all the requirements.

For developers who are deploying a high-speed, low-latency application and thinking about their storage requirements, it is useful to compare the two approaches.

To set the context, let’s consider a typical use case that we see here at Aerospike.

·  500 million objects

·  2 KB average size per object

·  50,000 transactions per second

·  A balance of 80% reads / 20% writes

·  A replication factor of two: one primary, one for failover

For these requirements a typical a typical in-memory database implementation using RAM for storage might look like this:

·  14 nodes

·  Two 2.4+ GHz CPUs

·  256 GB RAM

·  2 TB of free hard disk for persistence of the data in RAM

By contrast, by using SSDs for storage, we have been able to support the same use case with the following:

·  3 nodes

·  One 2.4+ GHz CPU

·  48 GB RAM

·  4 x 512 GB SSDs

It is worth noting that this configuration covers any SSD that has passed our 3x certification, which we measure through our open source Aerospike Certification Tool (ACT). Developers can view our test results for SSDs, or use ACT to conduct their own ACT benchmark testing.

By comparing RAM and SSD storage, the savings in initial hardware costs that SSDs enable immediately become obvious. Additionally, SSD storage results in further savings through the ease of management and reduction in power requirements.

For any developer who has thought about using a RAM-based storage solution, but has been daunted by the hardware, it is worth taking a look at the Aerospike Capacity Planning Guide to compare the benefits of using SSDs.

Young Paik is director of sales engineering at real-time database company Aerospike, Inc. He brings more than 15 years experience as a systems architect and engineer for enterprise software companies. For more information, visit, www.aerospike.com.

Download the whitepaper, Moving From Relational to NoSQL: How to Get Started. We’ll take you step by step through your first NoSQL project.

Topics:

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}