FIGURE 1: Oil and gas companies are using edge computing to predict and alert on malfunctions and failures
before they happen, saving millions in lost production yearly.
30 I ICT TODAY
use sensors ensuring that if a pump is about to have
enough of a problem that it shuts down the system,
and it warns someone first (Figure 1). Pumps that are
not running cost oil companies on average $40 million
per year with some estimates pegged at $80 million.
Then there is the question of the data. It used
to be common to install these little sensors, which
are basically faucets that spew out terabytes of data
every single day, but a lot of important information
just gets thrown away. Now organizations are beginning
to realize the tremendous unlocked value of this data
through machine learning. Machine learning, once
rooted in science fiction and thought to always be
just out of grasp, has started doing things that people
once thought were impossible for computers to do.
The intelligent building is turning into an artificial
intelligence building.
Edge compute is the practice of running servers
closer to where they are needed and closer to where
the gushers of data are produced. These sources of
data are not in the public cloud or some centralized
data center. They are attached to light fixtures
throughout a corporate office. They are attached to
the pumps on massive oil rigs off the coast of Africa.
They are sitting next to a jet engine on a 747. They
are even in a local supermarket ensuring applications
as mundane as a refrigerator not breaking down, but
if it does someone is notified before the produce goes
bad and must be thrown out. Grocery stores lose
millions of dollars a year on spoilage; some chains
estimate it to be 20-30 percent of sales! Companies
are not just installing sensors on these devices because
it is technically interesting. There is real business
value that is being created and captured. Oil companies