CTO Straight Talk - Issue 4 - 9

along the best path at any given moment.

name suggests, they don't process data;

However, this design choice also slows data

they only deliver it.

down, just as a multi-leg journey is slower
than an express trip. Even in the unlikely

As cloud computing emerged,

event that the entire Internet could be

companies have increasingly opted to

rebuilt to add non-stop connections from

migrate from on-premise infrastructure

cloud servers to end devices, simple physics

to hosted servers, moving data back

keeps even the most advanced ﬁber optic

and forth across the Internet. Despite

cables from transmitting it faster than the

the pervasive belief that cloud would

speed of light.

enable almost limitless scalability and
ﬂexibility, a 2009 article in the journal

" The two or three seconds it takes for

IEEE Pervasive Computing pinpointed a

data to travel hundreds or thousands

looming problem.

of miles is nothing for humans, but at
machine scale, it's glacial, " says Matthew

" We saw immediately that even though

Triﬁro, Chief Marketing Officer at data

centralization in the cloud brings about

infrastructure company Vapor IO

dramatic economies of scale, it would

and co-chair of LF Edge, an umbrella

also create huge issues with latency

organization under the Linux Foundation

and scalability, even before the huge

that aims to establish an open-source,

ﬂood of data becoming available from

interoperable, vendor-neutral framework

the Internet of Things, " says co-author

for edge computing. " That's because you

Mahadev Satyanarayanan, Carnegie

can't defeat the basic mechanisms that

Professor of Computer Science at

make the Internet work. "

Carnegie Mellon University. " It turns out
we wrote the founding manifesto of edge

That said, researchers have been trying

computing. "

for decades to solve the congestion
and latency problems inherent in the

" The Case for VM-Based Cloudlets in

Internet's basic structure. In the late

Mobile Computing " argued, presciently,

1990s, Akamai introduced the content

that concentrating most computing

delivery network, or CDN, a large
network of server nodes distributed on
the edge of the network near end users.
A CDN speeds up browser refreshes
and stream requests by prefetching
and caching high-demand web content.
Some CDNs also perform simple content
customization, such as inserting locationrelevant advertising. However, as their

1997

Th
These
companies don't want
to send their data traveling
to distant cloud data centers
and back before it can be
consumed.

1998

Computer scientist Brian Noble demonstrates how

MIT researchers incorporate as Akamai

to enable speech performance on a mobile device

Technologies to commercialize their algorithms

by offloading the actual data processing to a nearby

for intelligently routing and replicating content

server - the direct conceptual ancestor of Apple's

over a large network of distributed servers, which

Siri and Google's speech recognition services.

they call Content Development Networks (CDNs).
9
CTO Straight Talk - Issue 4

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