April/May/June 2020 I 33
necessary information across all stores in a consolidated
area to provide advanced services to their customers.
The client discovers glitches and errors due to the current
edge being the central point of processing for each
store independently which also requires a secondary process
to have the stores share information. Network managers
realize that in order to get latency to the desired level,
they need to get even closer to the source. A decision is
made to repurpose an existing piece of property in North
Carolina and build out a new, wholly owned, regional EDC
within a few miles of four nearby stores. The southern
Virginia multi-tenant edge is still used as a fog data center
for the critical data that can withstand the level
of latency provided.
A few years later, after losing hundreds of thousands
of dollars in frozen food due to refrigeration and compressor
failures, a decision is made to invest in machine
learning equipment that listens to the equipment operate
and detects patterns that suggest impending failures so
that they can be preventatively avoided. At the same
time, the client decides to implement a real-time automatic
checkout system that when customers remove
an item from the shelf and place it in the cart, the cart
tallies the bill, and their credit cards are charged at checkout.
These new systems have even more stringent latency
and bandwidth requirements, especially the mission critical
machine learning failure detection system, and it
is determined that even the round trip to the local edge
is not good enough. They determine that the most cost
efficient way to solve the problem without repurposing
valuable retail space is to purchase an unassuming shipping
container and within it a self-contained micro EDC
that is placed on the side of the building. This provides the
latency needed and allows the client to free up bandwidth
to its edge and begin selling services to other businesses
to utilize the excess capacity in its regional edge.
To address the many concerns in the industry about
security of an outdoor MDC, the unit could have been
placed indoors had the supermarket had a small vacant
office or closet that could be modified to include strict
access control, especially since sensitive information, such
as customer credit card accounts, would be at risk.
“Rob Hirschfeld, CEO of data center consulting company
RackN, said that an unmanned mini data center for
edge computing might make sense in parts of the US, but
leaving $100,000 worth of computing equipment without
some kind of guard is untenable in places like India.”5
ADDRESSING MDC CONCERNS
Some members in the TIA Edge Data Center Working
Group voice additional concerns about MDCs, which
include physical security, vulnerability to sabotage, no
on hand technical staff for troubleshooting resolution
and to service hardware quickly, exposure to harsh conditions
that may require hardened servers and switches,
and natural disasters like severe storms and hurricanes
that frequent the North Carolina coast in the supermarket
example. Currently, the TIA working group is developing
a list of edge use cases and analyzing effects on latency.
A partial sample list follows:
• LOWER LATENCY
• Smart factory
• AR/VR
• REDUCE NETWORK COSTS
• Local video processing for security and quality control
• Local distribution of content
• IMPROVE FORM FACTOR OF END DEVICES
• Less compute and memory requirements in head gear
(VR/AR), mobile cobots
• SECURITY
• Less access points, physical control
• HIGHER RELIABILITY
• Less failure points, less choking points
• TELECOMMUNICATIONS/ICT NETWORK SITES
• SDN infrastructure
• VEHICLE TO INFRASTRUCTURE V2I RSU
• Real-time traffic monitoring and alerts
• HD real-time maps
• V2V blind spot coverage
• NETWORK FUNCTION VIRTUALIZATION
• vRAN—Remote distributed unit (DU) and
centralized unit (CU)