Test software’s fourth wave

The application of software to testing is about to experience its fourth wave. Before we get to that stage, I’ll define the three previous waves as I see them.

In the first wave, there was the analysis of test data using computers. Early on, when engineers first got access to computing, typically in the form of minicomputers, they could gather their test data and enter it into the computer, typically with a punched deck of Hollerith cards, to run Fortran routines to analyze the measurement results.

Next came the use of stored programs in external computers that could automate the actual performance of tests. The stored programs controlled instruments that generated the test signals and those that captured the results, and the results were fed directly to the external computer for analysis.

In the third wave, the instruments themselves became computers, having microchips that controlled internal operations as well as taking care of the user interface and data storage. The instruments themselves could perform complex analyses of measurements without necessarily requiring an external computer, and could control other instruments. This evolved to a point where the definition of what functions an instrument performed was determined by software, dubbed a virtual instrument.

The fourth wave that’s about to hit is the need to deal with extremely large amounts of measurement data. National Instruments, the originator of virtual instruments, addresses this issue in a recent white paper NI Trend Watch 2014 , in which it refers to the issue by a term it has trademarked: Big Analog Data. (NI does not use the acronym BAD for this, but in the slang sense, it is “bad,” or good.)

NI cites many examples to show the vast amount of measurement data we are going to have to deal with in the near future, and makes the point that this data is intrinsically different than other kinds of big data. “Big Analog Data information is a little different from other big data, such as that derived in IT systems or social data. It includes analog data on voltage, pressure, acceleration, vibration, temperature, sound, and so on from the physical world. Big Analog Data sources are generated from the environment, nature, people, and electrical and mechanical machines. In addition, it’s the fastest of all big data since analog signals are generally continuous waveforms that require digitizing at rates as fast as tens of gigahertz, often at large bit widths. And, it’s the biggest type because this kind of information is constantly generated from natural and man-made sources. Consider the unceasing light, sound, motion, and electromagnetic waves throughout the world, solar system, and universe.”

Big Analog Data challenges include sensors and actuators, DAQ and analysis
systems, and IT infrastructures. (Courtesy of National Insturments )

To deal with BAD, NI says that engineers are seeking new, end-to-end integrated solutions for three-tier solution architectures (see figure). These solutions must that add insight from the real-time capture at the sensors to the analytics at the back-end IT infrastructures. Through the stages of data flow, the growing field of big data analytics is generating never-before-seen insights. Throughout tiers 2 and 3, data visualization products and technologies help realize the benefits of the acquired information.

 
NI also notes that reliability, availability, serviceability, and manageability (RASM) characteristics of these systems, which express their robustness related to how well they perform their intended functions, are becoming more important. They have a great impact on both technical and business outcomes.

NI concludes that Big Analog Data harbors great scientific, engineering, and business insight. “To tap this vast resource, developers are turning to solutions powered by tools and platforms that integrate well with each other and with a wide range of other partners.”