Dependability for Computer Systems meets Data Analytics
We live in a data-driven world as everyone around has been telling us of late. Everything is generating data, sometimes volumes of it, from the sensors embedded in our physical spaces to the large number of machines in data centers which are being monitored for a wide variety of metrics. The question that we pose is: Can the volume of data be used for improving the dependability of computing systems?
Dependability is simply the property that the system continues to provide its functionality despite the introduction of faults, either accidental faults (design defects, environmental effects, etc.) or maliciously introduced faults (security attacks, either external or internal). The computing systems that we target have been increasing in scale, both in terms of the number of executing elements and the amount of data that they need to process. For example, a large number of data-spewing sensors on mobile and embedded devices coupled with the large number of such devices show such increases in scale. We have been addressing the dependability challenge through large-scale data analytics in three broad domains: embedded and mobile networks, scientific computing clusters and applications, and computational genomics. In this talk, I will first give a high-level view of the dependability challenges in these three domains, how data analytics has been brought to bear on these challenges, and some of our key results. I will then go into two recent developments: dependability in a cellular network and dependability through approximating computation. In the first development, we answer the question – can the cellular network and the smart mobile devices working together mitigate the problem of network outages or reduced data bandwidth. In the second development, we answer the question – can the limitations of human perception be leveraged to approximate certain computation and thus allow the computation to meet timing guarantees, even when executing on resource-constrained platforms. A common example is video processing where the human visual system is forgiving for certain kinds of inaccuracy. I will conclude with some insights about how the power of data analytics can help us create more dependable systems.