The process of wrangling big data is fraught with pitfalls for enterprises. Data-driven enterprises are buckling under the burden of gathering, analyzing and making actionable an incredible and growing amount of data flowing in from a variety of sources. It's not just the amount of big data that is confounding data-driven companies: The speed at which data must be collected and analyzed, and the variety of data types (think: IoT sensors, log files, web clickstreams) are overwhelming enterprise data architectures, which are increasingly defined by a complex tangle of big data sources and processing systems. Topping all this is the problem of data drift, the unexpected changes that consistently plague big data sources and result in corrupt and unusable data.
In short, the complexity of data in motion is growing and risks undermining the success of the modern data-driven enterprise. A recent survey of data engineers and architects, conducted by StreamSets, sought to bring some perspective to the new reality in the enterprise, leading to some interesting insights about the enterprise data landscape.
As we expected, use of streaming data has become quite common, with a high number of respondents — 72 percent — collecting this data for a variety of uses. Of these, two-thirds (48 percent) collect a combination of batch and streaming data, since real-time data requires context to provide intelligence. In contrast, 28 percent move batch data only, and 24 percent ingest streaming data only.
Survey results also showed that enterprises are gathering data from a range of sources: 61 percent collect from transactional databases, 53 percent from log files, 42 percent from analytics databases, 27 percent from clickstream data and 18 percent from IoT devices.
Moving on from their use of streaming data, the survey reveals that enterprises are also experiencing a sense of data urgency — that is, expeditious analysis of their incoming data sets. In fact, according to the survey, 56 percent of respondents say they require data analysis within minutes of receiving the data, and 16 percent require analysis within seconds. The world has certainly evolved from the daily or weekly business intelligence report to a live dashboard, or even analysis that drives automated actions like website personalization that can have a direct impact on a business' effectiveness in engaging with its customers. These requirements put extreme pressure on enterprise data architectures not necessarily designed to deliver consumption-ready data with this type of speed.
Our survey responses indicate that enterprises funnel their data into a range of destinations, making them much more complicated and expensive to manage than ever before. In addition, respondents keep some of their data on premises (58 percent), some in private clouds (48 percent) and some in public clouds (27 percent). The combination of diverse data stores and multiple deployment models is a new phenomenon we call data sprawl, and it is a key driver of dataflow complexity.
The challenges of increased dataflow complexity are here and now and, given the unprecedented growth of data each day, must be considered the new normal. With this information as a bird's-eye view of the state of data in motion, savvy enterprises will adopt technologies and solutions that will help them evolve with the big data landscape.
Pat Patterson is Community Champion at StreamSets.