CTO Straight Talk - Issue 4 - 11

resources into a few enormous data

Between Tier 1 (the cloud) and Tiers 3

centers hundreds or thousands of

and 4 (the edge) lies Tier 2, where edge

miles apart would inherently hobble

computing takes place. Tier 2 receives

applications that are hungry for

data from nearby devices, processes

bandwidth and/or sensitive to latency

it, and pushes the results back to the

and jitter. The solution it proposed

devices, occasionally retaining some of

was to create many small but powerful

that data in short-term storage. The only

CDN-like server nodes at the edge of

data that passes from Tier 2 on to Tier

the Internet, to eliminate the need for

1 is data that doesn't have immediate

large amounts of data to travel to and

relevance but is useful in the long term,

from remote data centers before being

such as for deep analytics or customer

consumed.

management.

" At ﬁrst, we called them cloudlets, but

Once micro data centers are ubiquitous

you also see them referred to now as

- in parking lots, on cell towers, in

fog nodes or micro data centers, " says

wiring boxes, under park benches, on

Satyanarayanan, who has continued

fences - it will be easier to deliver low

reﬁning the concept into what is now

latency data in a massively scalable

a four-tiered infrastructure (see the

way, because there will always be a

graphic " Competitive Edge " ).

nearby node to process data to and
from phones, cars, drones, sensors, and

HOW EDGE
COMPUTING WORKS

other edge devices. As a bonus, edge

Tier 1 is the cloud or any other wide area

well before it can propagate to the rest

network, or WAN. Tiers 3 and 4 are the

of the network.

computing will also make it easier to
identify and isolate suspicious activity

" edge. " Tier 4 is made up of devices
such as cameras, sensors, and smart

Companies using edge computing

appliances that generate data but do

will be able to deliver the required

little or no processing. Tier 3 includes

information to end users at the speed

similar devices, such as smartphones

they expect, and in the process, they'll

and wearables, that have some compute

have more data from more sources-

power but can't extract the full value

with some that is less time-sensitive

of their data because of limiting factors

sent back to the cloud for analysis-to

such as weight, size, battery life, and

transform into a deeper understanding

heat dissipation.

of their customers and their market.

1999

2001

In February, Akamai delivers its ﬁrst live traffic:

VMware begins selling a software virtualization
rtualizatio
tu

a single pixel hidden deep within the Disney

solution that lets multiple operating syst
systems
tems rrun
n

website. Two months later, it launches commercial

simultaneously on servers. This makes
kes server
rve

service with Yahoo!, then one of the world's most

farms possible, enabling cloud computing.
mp ng

trafficked websites, as a charter customer.
11
CTO Straight Talk - Issue 4

Table of Contents for the Digital Edition of CTO Straight Talk - Issue 4
