14 - Cloud Computing [Massive Teaching]

From Week 2 Lecture Videos of "Teaching goes massive: new skills required"
by Paul-Olivier Dehaye
See
https://etherpad.mozilla.org/pr8ZtLXODg
and
http://www.elearnspace.org/blog/2014/07/09/congrats-to-paul-olivier-dehaye-massiveteaching/

Imagine you want to build a software-based
company, providing a service on the web.

A few years ago, it used to be that you
needed to do two things:

develop the software and buy
expensive hardware to run it on.

That's not true anymore.

Nowadays, you still have to develop the
software, of course,

but you don't need to buy the hardware.

You can rent it, and in ways that are
extremely flexible and match your needs.

This is what's called cloud computing.

So, what's a cloud?

It's an abstraction of a computer or many
computers.

Hmm, that doesn't really help,

so, let's start at the beginning.

What's a computer?

Essentially, it has two components that are
important:

Number one is memory, like a hard drive.

Number two is a processing unit, which
helps derive new results from old ones.

That's the unit that does all the
computation, the brain --

the brain if you want.

This used to only come packaged as a box,
with the brain and the memory together.

For instance, think of your laptop or your
desktop.

So let's look at a typical example,
such as the Coursera start-up,

and its options for growing.

At the beginning you just have a few students
using the platform,

which runs on one server.

Then that central processing unit and storage unit
-- so the brain and the memory --

will do everything, switching between
serving videos to different students,

serving pages, computing grades, searching
posts in the forums, etcetera.

This would work well,
but as soon as Coursera is successful,

it doesn't work anymore.

You have to buy new servers, keep up with
demand, and also keep everything balanced.

You have to make sure that you have
the right number of these machines,

at the right place.

And it's going to be expensive if you
under-use some parts of them.

Now with cloud computing, this is
essentially not a problem anymore.

You get very flexible computing units
and very flexible storage units.

They're flexible simultaneously in
geographical location and in size.

So you can ramp each,
depending on real needs.

For instance, in the evening, Coursera will
probably see higher loads

from most of the students doing their homework
and it can adapt easily to that.

Who provides those cloud computing services?

Well, more and more companies do that,

but the most famous is probably
Amazon Web Services.

This is a part of Amazon that was created

to answer their own needs with their
e-commerce website

that would see huge demands around
Christmas and so on.

Writing software that uses cloud computing
is not so hard,

but it's very hard to rewrite software
that was written

before these ideas took hold.

Older learning management systems, that
did not explore those ideas,

even if they were really good
at the services provided,

will probably need to be rewritten,

in order to scale to millions of students.

Not everything is perfect
with cloud computing, though.

Since the data is heavily distributed
around the world,

a whole lot of new problems can come,
relating to data privacy, for instance.

[CC BY-SA
Paul-Olivier Dehaye]