Network Speed
Lowering the Startup Barrier to Disruption Through
De-Centralization
By: Yermo Lamers
We have a strong natural bias to keep
using existing ideas that have served us well. Once established, changing our
thinking is difficult. But the world around us changes relentlessly. Here-in
are sewn the seeds of disruption.
Oftentimes, incremental and seemingly
insignificant changes in technology have huge effects that are not immediately
apparent. Network speeds, which have been incrementally improving for decades,
are such a change. Sure, we can do things faster, but what does it mean?
Human beings making a request of an
information system will typically start to get bored after a few seconds and
frustrated after a few more. That's the benchmark we look for in getting a
response. In the good 'ol days, network speeds were only fast enough to
transmit a few raw characters over any distance in that time-frame. I remember
as a little kid playing the game of Zork after hours using a Silent 700 thermal
paper terminal connected using an acoustic coupler modem to a minicomputer at
NASA. I think it could only transmit at 110 baud. You typed in a line, “Pick up
sword”. You waited a second or two and then the thing started whirring a
response back at you. That system could only transmit one line of characters at
a time “fast enough” to match the human expectation. It was a natural
consequence of these slow connection speeds that the world was ruled by dumb
terminals connected to centrally located mini-computers and, on the high end,
mainframes. Huge proprietary businesses with wide moats were built protected by
this centralization.
Then one day, I heard about this
impossibly fast new technology called Ethernet.
“10Mbits/s?” There are those who say the PC solely brought about the end
of minicomputer and mainframe era. I would disagree and suggest that it was, in
fact, Ethernet that was the key to disrupting their world. Ethernet was fast.
But what did it mean? Ethernet meant you could now inexpensively hook
commodity machines together to quickly distribute data in a way that was not
possible before. Expensive minicomputers which used to be data store and
computational powerhouse combined could now be replaced by commodity machines
that acted essentially as nothing but a data store. Computation was offloaded
to relatively inexpensive workstations. The world of client server was born and
a whole new industry came along to disrupt the one before. Importantly,
economies of scale made the knowledge to run these machines a commodity which
lowered one of the big barriers to starting new businesses. Namely, talent was
now available. In a way, it was increases in local area network speeds more
than the PC itself that enabled the rise of Microsoft. Microsoft was able able
to see ways to exploit the new context with fresh eyes, at IBM's expense, who
were still caught in thinking that the centralized models that had worked
before would continue to be competitive.
Microsoft built an empire on the local
area network. They controlled the server and the workstation. What I never saw
was how Microsoft's moat was tied to a particular Goldilocks zone of wide area
network speeds.
It started some time in '93. I remember
getting a US Robotics Dual Standard modem. It could talk to another modem of
the same type at 14,400. What did this mean? It meant I could download a
complete distribution of Linux in some reasonable time. As modem speeds
increased, it became easier and easier for programmers to start distributing
what they had written. The free software movement had been around for quite
some time, but it was the advent of the high speed modem that, in my opinion,
was key to it's rise. Just as I did, Microsoft failed to notice that this was a
harbinger of things to come. Linux was not the threat; not by itself. Network
speed was the real threat to it's business model.
As long as broadband speeds were low
enough and it was impractical to distribute truly large quantities of data quickly,
Microsoft's position was defensible. They still controlled the client and the
server on the LAN. However, at some point, wide area network speeds became fast
enough to disrupt Microsoft's stranglehold on business processes. Mired in old
ideas one might think this is because it was easier to distribute large
quantities of software thus threatening Microsoft's hold on distribution. I
would argue that once network speeds became fast enough sometime in the early
2000's, it led to the ascendancy of business models that could not have
succeeded before. The era of Google, Facebook and other web 2.0 companies was
upon us and Microsoft's model of services centralized to the LAN was suddenly
feeling antiquated. What did these new high speed network connections mean?
Third parties somewhere out on the vast internet could now deliver experiences
to users in the critical couple second window that rivaled the experiences
delivered by local desktop software. The browser became the platform. This is
actually what killed my little stock market software company. It turns out
users hate installing software. They hate updating it. They hate not being able
to use it where-ever when-ever they want to. With functionality delivered by
Web 2.0 there's nothing for the user to install or update. They just log in and
use the service on whatever device they want. The availability of these Web 2.0
services also had a positive effect on small business. It meant that the costs
to start a new business had once again been reduced. A business could now start
out without needing to run it's own administrative servers or even network.
Just use Google Docs, online payroll, and accounting service to get off the
ground. There little need to fund an expensive IT staff for internal operations
anymore let alone pay Microsoft's licensing fees. Additionally, these services
enabled a more mobile and distributed workforce. Most small businesses can't
afford the infrastructure costs to make their internal Microsoft dominated
networks available to a mobile workforce. Now with Web 2.0, they get a mobile
enabled work-force for free. This in turn enables the business to look for
talent where it happens to be regardless of whether it's local, across the
country or around the world. The local area network that Microsoft dominated
was another kind of centralized model disrupted by network speed increases.
Network speeds continue to increase and
there's a new potentially larger disruption of a centralized model in the
works. Network speeds are starting to go beyond “human response time” speeds
and are beginning to reach what we can call “machine speed”. Reaching a speed where a rich experience can
be delivered to a user within their boredom threshold was the catalyst to disrupt
one of the most successful businesses in history. What effect would it have if
network speed increased to the point where in-machine and inter-machine data
transfer rates become less distinguishable?
A core assumption in the last 40+ years
of operating system design is that operating with the outside world is slow.
Machines are distinct and services are centralized on the machine. I have my
machine and you have yours and they run distinct operating systems. Even if I
have a datacenter, each machine runs it's own OS.
If the network is so fast that I can
call services between machines at something close to “machine speed” across the
country or even around the world, it might imply the paradigm of single
distinct machines starts looking dated. What new disruptive decentralization
might occur?
Web 2.0 companies brought zero-install,
zero-maintenance services to the people side of the business. For many
companies, especially heavily online enterprises, there are still very
significant costs associated with providing the core services of the business
to it's users. There are servers to configure. There are programs to write and
third party components to integrate. There's data base administration to do and
patches to apply. There's also excess capacity which has to be built,
maintained, paid for and left idle to handle the occasional unexpected surges
in usage. Then there are all the very expensive salaries to pay. I've heard
many entrepreneurs bemoan how difficult and expensive it is to find talent.
Imagine being able to launch a new initiative quicker, with less overhead, less
staff and less need for development and administration.
This new disruptive force is called the
“cloud”. And by the cloud, I do not mean just virtual servers that you still
have to administer. I mean the ability to distribute and auto-scale the
components that have traditionally made up online software systems. This is
called Platform as a Service and it does for software what Web 2.0 did for
people. It radically decentralizes it.
With “Platform as a Service” (PAS)
models, you no longer run any kind of server at all. You develop your
application which represents your unique value proposition and then push it out
to your “PAS Cloud” vendor. The vendor runs your application as just another
process in their distributed network of machines. There's no administration and
there is built in scalability. If you happen to get favorable press and
suddenly need more capacity there's no need to wake up your IT staff, since you
don't have one. You simply open your control panel and spool up additional
instances of your application to match demand. There's no excess capacity for
you to carry and pay for unnecessarily. Your operational costs are reduced.
Your ability to respond is increased.
But more importantly, there is an
ecosystem of third party online service components rapidly developing which are
likely to shorten online software development cycles. There are features almost
every online application requires such as image manipulation, databases, messaging, e-commerce, tracking,
reporting, and alerting available among many others. In the past, these
services had to be developed or at least downloaded and installed on some
server. Then they had to be maintained, patched and upgraded. Imagine, if such
services were simply available online somewhere that you could simply sign up
for and for nominal incremental cost just hook into your application and start
using. Imagine that these components could scale based on demand. Sure,
resizing a couple hundred images is no problem. But what if you run into
hockey-stick adoption and need to scale a million tomorrow? Imagine how the act
of building applications could become nothing more than hooking up distributed
pre-built scalable zero-administration
services together.
The context has changed. Old ideas
based on centralization are a competitive disadvantage. In this new context of
faster network speeds, the cloud and Platform as a Service models, new
initiatives can be built with less funding, less staff, less infrastructure and
with a much shorter time to market while being more scalable and vastly more
distributed.
What will it mean when networks become
faster still? What kinds of hidden centralization that we don't even question
now will be disrupted?