If you want to support a mobile app environment, think carefully about your approach to data analytics. Monolithic...
analytics applications are limited to pregenerated reports that lack specificity or timeliness. They run independently and generate queries against one or more databases. These limited query options and generalized reporting features often fail to deliver the granularity and insight mobile end users expect.
However, distributed analytics help spread data analysis workloads across nodes in a cluster of services. Because of this, distributed computing can provide more support for the custom algorithms and flexible visualization techniques that today's mobile end users require.
The challenge of analytics
Today's data-rich environments draw information from many sources -- a fact that highlights the drawbacks of limited queries. Stand-alone, monolithic approaches are often slow, expensive and rife with integration hurdles. Also, consider that, once you install an analytics application, you still need an IT team or a third-party provider to procure that data for the application. This only adds challenges in a time when business processes change constantly and organizations are pummeled with new data sources.
These challenges explain why analytics and business intelligence (BI) deployments were limited for many organizations. However, massively parallel processing data warehouses and SQL-on-Hadoop systems now offer fast and inexpensive resources that help organizations analyze data within the database and eliminate extract, transform and load functions. These resources offer a query-at-will framework in which you can retrieve a wide range of data.
The value of distribution
When multiple nodes work on the same data, they provide insight faster than a single node with linear processing can. Moreover, today's businesses often store a diverse range of data source types. This requires teams to create a transformational layer before reporting happens. Increased numbers of organizations use Hadoop Distributed File System (HDFS) as a repository for these large amounts of data.
A key attribute of HDFS is that it spreads data over different clusters, which enables relatively easy integration for distributed analytics applications. Today, organizations generally work with a combination of relational data, real-time data and historical data. This data could reside on premises or across a combination of public and private clouds.
It only takes a single analytical application tool to integrate and govern that data. This means that teams can create a dynamic model that consistently responds to end-user requests with the correct analytical query. The viewing options for this data include real-time interactive visualizations for data scientists, dashboards for performance metrics and embedded visual analytics that serve mobile end users.
The value of microservices
Combining BI and analytics software with a microservices approach enables average end users to drill down into data with specific types of queries. When it comes time to visualize that data, organizations must decide whether to build customized visualization tools in-house or adopt a third-party option.
A vast number of options exist for visualization, which include web-based platforms and stand-alone, open source tools. These tools tend to focus on a range of data interaction, from complex depictions of near-time data to simple renderings.
However, big data sources have their limitations. Streaming and unstructured data sources present challenges that mainstream analytical tools struggle to depict. For example, some query connections won't accept data set blending, which limits exploratory analysis. Teams may also encounter system timeouts, out-of-memory exceptions, long query waits and rendering limitations.
However, distributed analytics approaches can excel in big data. For example, Zoomdata offers a microquery approach that can directly connect to the broadest sets of big data sources for analysis and rendering without the need to move the data. Zoomdata quickly provides both visual and interactive depictions of big data through its patented data sharpening and microqueries.
The responsibility of IT
For business end users, the number of available data visualization options can be overwhelming. It's important for IT leaders to clarify the organization's goals surrounding data analysis and identify the available query types, skill levels and technical resources. It's also useful to record both the quantity and quality of the target data. Organizations that solve these matters will become well-suited to select the most appropriate tool for the task.