FIGURE 4: Enabling mainstream autonomous vehicles may require thousands of sensors monitoring and reacting to every aspect
of the surrounding environment requiring instantaneous processing and feedback with little or no tolerable latency.
April/May/June 2020 I 31
Another ultra-low latency application of interest is
autonomous vehicles. As they become more mainstream,
there will ultimately be thousands of self-driving vehicles
in a given town or city that must provide awareness of
other cars, detection of bikes, animals, and people to avoid
accidents while human occupants have their noses buried
in their devices making that all important text, reading
road conditions, traffic prompters, and changing conditions,
as well as paying attention to operational and system
failure mechanical monitoring (Figure 4). Due to the
vast number of sensors and processors, some people may
envision the need for thousands of EDCs in the form of
prefabricated MDCs located a couple of blocks apart.
Some members in the TIA Edge Data Center Working
Group, on the other hand, do not see autonomous vehicles
requiring a multitude of EDCs as these vehicles have
on-board intelligence to make detections and decisions.
They only require a centralized data center for things like
maps and traffic information, which are not delay sensitive,
since maps can be cached locally or downloaded
at the beginning of a trip. For traffic lights, managing
congestion, and gathering data, the municipal or local
city data center or a nearby colocation facility either acting
as an EDC or handling edge applications are sufficient
and efficient for current smart city deployments.
EDCS, GEOGRAPHY, AND ITS
CONTINUAL EVOLUTION
Much of the general data that is stored or processed
today can withstand latency times between 10 ms and
100 ms, including some popular video streaming services.
In this case, large scale core data centers are adequate.
Core data centers are geographically dispersed and cover
large regions. In many cases, there may be a path to two
or more core data centers. This means that the data may
go to one location, such as Ashburn, Virginia or Atlanta,
Georgia, or it may make a round trip touching both or
may even make multiple hops. When latency requirements
lower to around 1 ms, higher bandwidth dictates,
and there may be a need to move the process function