When is NoSQL the Right Database Choice?

As unstructured data surpasses structured data in many enterprises, interest in NoSQL is increasing as well because of its ability to shed light on the nuggets of valuable information buried in dark data.

This unstructured data is rushing in from an array of new sources, noted Unisphere Research analyst Joe McKendrick in a recent DBTA webcast, titled "When NoSQL is the Right Database for the Job." First, there is the machine-generated data – log data, application to application data, geospatial data, data from smart meters, manufacturing sites, healthcare facilities and distribution companies. In addition to the data generated by this so-called Internet of Things, there is the massive volume of data being generated by people in the form of word documents spreadsheets, social media interaction, plus data included in systems such as monitoring, audit and security data.

Creating an Enterprise Architecture that Can Handle Unstructured and Structured Data

According to McKendrick, creating an architecture that can handle all that data along with traditional data types handled by relational database management systems is on the horizon for many organizations. The key is to enable an environment of co-existence between the relational database systems and unstructured data environments.

It is also important to remember that no approach is the right one and that every organization has different data requirements. For some, that may mean bringing legacy infrastructure into the fold, while for those starting up new operations, integrated green field environments developed from scratch may be possible, McKendrick said.

NoSQL Means Expanded Opportunities for Data Analysis

NoSQL databases are growing increasingly popular, noted McKendrick, because they offer low cost, low complexity solutions to capturing and analyzing unstructured forms of data in a way that was not available until recently. For example, if you look at RFID tags and machine data, that has existed for many years now, but there has never before been a cheap or efficient way to do anything with the data from these source that was being amassed. Now, there are much more inexpensive databases and tools.

NoSQL is Expanding in Market Acceptance

McKendrick cited industry research showing that the worldwide NoSQL market is expected to reach 3.4 billion by 2018.

He added that while the great majority of organizations still rely on relational technology, according to a 2013 Unisphere Research study, large segments are also adopting NoSQL. Big data technologies adopted or soon to be adopted include Hadoop/MapReduce (28%), other open source technologies (20%), NoSQL databases (20%), and R (12%).

NoSQL Offers Strong Advantage for Key Use Cases

NoSQL offers particular advantages for specific use cases, pointed out David Segleau, director of product marketing at Oracle. Webscale transaction and personalization is the ‘poster child’ for NoSQL, he said, while real-time event processing and sensor data management are other key use cases.  

When considering a NoSQL technology, it is important to make the right choice early on since no one has the luxury of time for experimentation. Having the ability to integrate with the rest of the IT environment is critical concern for organizations, he added, pointing out that if this is not addressed early on, it is a price that will be paid over and over again. Reliability and supportability are also imporant considerations. And, it is also important to work with a product that behaves predictably and is supported by a company that stands behind it.

Segleau highlighted some of the advantages of the Oracle NoSQL Database, including the fact that it integrates with the Oracle stack, is supported by Oracle, both in the enterprise edition and the community version, and is complementary to the Oracle RDBMS.

Where Relational May Fit Better

While offering significant advantages for a variety of enterprise needs, clearly, NoSQL is not designed to replace relational, Segleau emphasized.

Relational technologies may be the best fit, said McKendrick, if SQL statements or more complex queries are needed; if the business case for unstructured data is still being defined; and if rock solid transaction consistency (ACID) is required. Additional considerations include the maturity and security of the system, support of the vendor, and support for analytics/BI and reporting tools.

To replay the webcast, titled "When NoSQL is the Right Database for the Job," as well as audience questions and answers, go here.

Image of hand writing dark data with marker courtesy of Shutterstock.