All things Big Data, Hadoop, and NoSQL

Dale Kim

Subscribe to Dale Kim: eMailAlertsEmail Alerts
Get Dale Kim: homepageHomepage mobileMobile rssRSS facebookFacebook twitterTwitter linkedinLinkedIn

Related Topics: Enterprise Architecture, Big Data on Ulitzer


Data Lakes: Building the Business Case By @MapR | @CloudExpo [#BigData]

The data lake will have a profound impact on enterprise data architectures

Data lakes are the business buzzword of the day. The excitement around this new way of working with Big Data is more than justified, as the vision of a consolidated analytics platform for important business data is compelling. But the lack of a clear set of implementation strategies and repeatable business value is causing some hesitation. It is a phenomenon that we also saw when concepts such as "Big Data" and even "data warehouse" first emerged in the industry.

As did its aforementioned predecessors, the data lake will have a profound impact on enterprise data architectures, enabling us to work with unstructured data in its native format, in addition to data in specific structures, as is necessary for data warehousing. Leveraging a native format helps to maintain data provenance and fidelity more easily, and allows a myriad of analyses to be performed using the same data. Using a Hadoop architecture, which enables distributed big data processing utilizing open software standards and massively parallel commodity hardware, brings additional benefits around cost and scalability to the data lake.

Solutions... and Challenges
Data lakes solve the data integration challenge that many businesses face as they incorporate new data sources, including mobile and cloud applications, into their enterprise data architecture. They also help with analysis of the ever-increasing flow of information generated by the Internet of Things. Companies typically capture the data for immediate analytics but can also store the rapidly growing volumes of data for to-be-determined future uses.

A single enterprise-wide repository of data provides nearly unlimited potential for actionable information and data discovery, now and into the future. As volume, variety, and velocity continue to grow, so do the business intelligence benefits.

A new world of data exploration opens up for a variety of end users as well. The standard relational data warehouse solution requires significant reliance on IT, whereas a data lake can be accessed in a variety of ways without waiting for IT to prepare the data.

Given the many obvious benefits, why do some lake initiatives fail to thrive? Typically, lack of planning tied to specific business initiatives. If a company simply dumps data into the lake as merely a central repository, with no plans around business metrics, the lake can easily become a data graveyard.

The Data Lake Adoption Process
Data lake adoption in your enterprise should be approached by a series of stages. The first stage often comprises proof of concept projects that focus on particular business outcomes. This process serves as an introduction to using and managing a data lake, and perhaps the Apache Hadoop environment if Hadoop is a new addition to a particular business's ecosystem. Determining one or two pilot projects, particularly those with measurable results and real value to the business, is an excellent way to launch a data lake, and the best way to circumvent the data graveyard problem.

For the next stage, one common pattern, as noted in The Definitive Guide to the Data Lake, a white paper by strategic research firm Radiant Advisors, is to leverage Hadoop's scalable, low-cost persistence layer or its ability to perform big data processing and analytics. An example of this would be relocating historical warehouse data into the data lake as long-tail analytics engine to reduce the overall size and management of the data warehouse.

The next stage often features initiatives to further consolidate data for big data and analytics projects. Additionally, companies will focus on leveraging Hadoop's affordable scalability to increasingly capture data from disparate sources such as social networking platforms, the Internet of Things, and other sources of unstructured data.

The final stage, stage 4, as described in The Definitive Guide to the Data Lake, is that of continuous optimization. "Inevitably, there comes a tipping point when a data lake becomes a core part of IT's strategy and planning, and walls between operational workloads and analytic workloads are rethought as a single enterprise platform that leverages the strengths of each architecture component.  Consider Stage 4 to be the level at which the journey to the data lake has Hadoop fulfilling a foundational component of the enterprise data architecture strategy, and supporting more of the operational, analytic, and big data workloads with both persistence and data engine layers," the Definitive Guide notes.

Managing the Data Lake
A data lake indisputably enables businesses to build an efficient data architecture that can meet the needs of various enterprise application workloads, and which will gain efficiency through the reusability and consistency of data. However, the rate at which any company can accelerate data lake adoption will be affected by the establishment of robust data governance, security, roles, and access policies.

For more information on best practices for managing/governing and securing a data lake, along with a detailed explanation of the critical factors that typically determine the success of any data lake project, download The Definitive Guide to the Data Lake by Radiant Advisors.

More Stories By Dale Kim

Dale is Director of Industry Solutions at MapR. His technical and managerial experience includes work with relational databases, as well as non-relational data in the areas of search, content management, and NoSQL. Dale holds an MBA from Santa Clara University, and a BA in Computer Science from the UC Berkeley.