This content is part of the Conference Coverage: OpenWorld 2017 in review: Oracle cloud offerings at center stage

A transition from Oracle to NoSQL Cassandra

Several companies are augmenting their Oracle database workloads with NoSQL Cassandra. Why?

LAS VEGAS -- Oracle databases might be old and steady, but for some companies, it's time to switch things up.

Take PayPal Inc. After some hard examination, PayPal determined the Oracle database is no longer a right fit, and the company needed a change.

Enter the NoSQL Cassandra database, which has quickly evolved as a challenger to the Oracle relational database management system (RDBMS).

PayPal, a worldwide online payment system based in San Jose, Calif., is one of several well-known major companies that use Cassandra to augment and potentially replace traditional Oracle workloads. Other companies include Wal-Mart, eBay, Netflix, Macy's, Target, ING, Capital One and Intuit.

"We at PayPal use a polyglot of database technologies," said John Kanagaraj, architect of core data platform at PayPal, here at the Collaborate 17 conference this month. "These include Oracle and NoSQL databases, such as Cassandra, Couchbase, Aerospike, MongoDB and Hadoop."

PayPal started its NoSQL journey in late 2012, and the company has multiple use cases for it, including low-latency caching, a cookie and token store, and journaling, for which Cassandra is especially useful.

The primary challenges at scale for PayPal, Kanagaraj said, were connections, memory, interconnect, data definition language on busy tables and Real Application Clusters configuration.

"The solutions were replication, connection multiplexing, custom caching, sharding and, now, active-active scale-out using NoSQL," Kanagaraj said.

So, why the NoSQL Cassandra database?

As Kanagaraj explained, RDBMS databases are essentially ACID (atomicity, consistency, isolation and durability), which can inhibit scalability. RDBMS databases also have a lack of native sharding and a higher base cost for setup and scale.

On the other hand, NoSQL databases, such as Cassandra, can scale highly and quickly at a lower cost, have a low-latency and scalable K-V read/write, a flexible data model, and an open source and enterprise model.

Cassandra essentially operates as a column database, with key values included -- similar to a SQL database, with the idea of grouping tables and replication. For a company such as PayPal, Cassandra is especially useful because it manages large-scale volumes of data.

[NoSQL Cassandra] provides a rapid web-scale growth for external entities and users.
John Kanagarajarchitect of core data platform, PayPal Inc.

"It provides a rapid web-scale growth for external entities and users," Kanagaraj said. "It also has an ability to support a burst of traffic patterns."

There are advantages to RDBMS and potential pitfalls to a Cassandra NoSQL database, however.

For RDBMS, ACID is essential in many cases, Kanagaraj said, and it also supports a complex data model. Oracle databases are also a mature technology and ecosystem, and they contain a wider skill availability, as more employees are proficient in it.

For Cassandra and other NoSQL databases, the limited ACID could present problems, and the technology is still evolving, so there is a maturing but fragmented landscape. And skill sets are not widely available -- but that's coming, Kanagaraj said.

PayPal decided the pros outweigh the cons with NoSQL, especially concerning open source.

"Almost all NoSQL products are open source, and it is relatively quick learning," Kanagaraj said. "There are quick and easy installs, and community versions from vendors."

In addition, he said, NoSQL Cassandra databases cater to unstructured needs and primarily solve an impedance mismatch.

Next Steps

Compare MongoDB vs. Cassandra databases

Upgrade Oracle EBS 12.2.5

A NoSQL database guide

Dig Deeper on Oracle MySQL database