edfas.org
ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 18 NO. 3
20
and three open/close sensors. In addition, I usemore than
20 apps in my home.
So many devices and so many apps unquestionably
bring me a great deal of personal benefit; however, so
many devices and apps create issues, too. The biggest
challenge I face in my personal “IoT@Home” is that I
constantly need to fix one thing or another. Each device
has an approximate average failure rate of once per year.
So, if just one device fails per week, that leaves me in a
regularmode of debuggingmy system, and that keepsme
really, really unhappy.
Device failures currently comprise a problemwith only
100 devices in my home, but what happens when the
forecasts for IoT proliferation come to pass and I have 300
or 400 connected devices in my home? Most home users
simply are not interested or are incapable of dealing with
each individual fault across devices in a system that is so
quickly growing in complexity, interconnectivity, and the
sheer number of devices.
TOWARD A SOFTWARE-DEFINED IoT
Implicit in the issue of device failures are at least a
couple of calls toaction for the industry that is building out
the IoT 1.0 around the world. Device manufacturers can
strive tomakemore-reliable devices, and, of course, they
already are and always will be striving to do so. Certainly,
this is a necessary pursuit.
However, even if devices are made more reliable, the
truth is that some rate of device failure is inevitable. The
IoT 1.0 will deliver its greatest benefits when users are
able to experience the IoTwhile remaining almost naïve to
the applications and devices that enable the experience.
The system should still be able to operate, even when
individual devices run into a faulty state.
This need tees up the requirement for an intelligent
middleware tominimize human effort and automatically
monitor and control the overall system, recognize indi-
vidual failures, and hand over capabilities among devices
as necessary. There needs to be an intelligent mapping of
devices within the virtual space of the IoT 1.0. This cyber-
physical intersection will be critical to a resilient system.
At the Intel-NTUConnectedContext ComputingCenter,
for example, we are working on a proactive management
framework, “WuKong,” that works to limit IoT user inter-
action to simply sending requests to applications and
defining context and high-level policy. The new middle-
ware layer intelligently maps the logical relationship to
physical devices, and, when devices fail or are replaced,
themiddleware automatically re-maps a logical relation-
ship to the physical devices (Fig. 3).
Furthermore, programming is performedonhigh-level,
hardware-independent construction—not the specific
physical devices—so that programs can be written once
and then run everywhere across the IoT. Moving the
devices/services around becomes easier. Finally, if the
user’s intention is properly communicated to themiddle-
ware, then the user should not be concerned with finely
adjusting the device sensitivities.
Sensitivity is another area of pain. For example, today
a motion sensor set too sensitive on my front porch may
detect a car passing on the street in front of my home
and trigger useless picture-taking by my home-security
system; the samemotion sensor set not sensitive enough
may ignore someone walking on my front yard. With a
better, more intelligent middleware that understands the
security goals of the user, the middleware should make
sensitivity decisions per device basedon the user’s greater
system-level intention, as opposed to simple, preset
thresholds for each device.
Such a middleware layer is being designed ultimately
to enable a software-defined IoT that would minimize
human intervention and relieve the pain point of manag-
ing devices. It is one of the places where the IoT demands
open, cross-discipline collaboration to rapidly and fully
bring about the benefits envisioned. I invite you to visit
http://iot.ieee.org/iot-scenarios.htmlto weigh in on the
Intel-NTUConnectedContext ComputingCenter’s concept
for intelligent IoT middleware and other emerging IoT
scenarios.
CONCLUSION
Collaboration is the key for a large system such as the
IoT 1.0 to function optimally. Many different components
across diverse application domainsmust be able to seam-
lessly interoperate, and each application domain has
insights that must be taken into account for the greatest
potential benefit of the IoT to be realized. Technologies
must and will advance so that the system is still able to
operate even in the reality of device failure.
Fig. 3
Middleware will enable a software-defined IoT,
minimizing human intervention.