Big Data comes in two distinct forms, at rest and in motion. The distinction does not lie in the structure (or lack of structure) of the data itself. Both kinds of Big Data can and probably do include both structured and unstructured data – typically with a lot more of the latter than traditional data management had to contend with.
The distinction, as the names imply, is in how the data is handled. Data "at rest" is not permanently at rest – sooner or later you will be bringing it up to analyze it and interpret it. But you won't be doing that until the end of the day, month, quarter, or whatever. Until that time the data can be warehoused in a relatively traditional way.
Big Data "in motion" is processed and analyzed on the fly – in real time, or nearly so. Which means that it has to be handled in a very different way than traditional stored data. Jack Vaughan at SearchDataManagement notes that the handling of Big Data in motion "tends to resemble event processing architectures, and focuses on real-time or operational intelligence applications."
Here you are not looking back at what happened last quarter or last month, or even yesterday: You are looking at what happened a minute ago – and what is happening right now. Which means a considerable added challenge to managing the data both safely and effectively.
Data storage engineers have made great strides in designing for Big Data, but these solutions are often intended primarily for data at rest. When used to manage Big Data in motion, these solutions can fall behind. Says Jay Kreps of LinkedIn, problems are likely to appear "where latency is the primary concern – where people are not just reporting day-to-day."
In short, if your company needs or wants to track and analyze Big Data in real time, it needs to adopt a data storage and management architecture that can handle data on the fly. On the flip side, if latency is not a problem, and data only needs to be examined retrospectively, investing in costly solutions for Big Data in motion is a needless expense.