NEW: British consultant and publisher Charles
Brett has a new free form of his journal, called INSIGHT-SPECTRA. I am pleased to tell you that I have an article,
Cloudosophy, in the first issue.
Although
it's a bit long for a blog post, I have included it here. You may
recognize some of the comments and charts from earlier blogs.
You can get a free subscription to
INSIGHT-SPECTRA by going to their site and filling out the form at the bottom of the page.
www.insight-spectra.com .
Brett’s former publication, SPECTRUM MIDDLEWARE
was a well-respected journal and I was honored to occasionally write for
it. I’m hoping to continue that
tradition in INSIGHT-SPECTRA.
CLOUDOSOPHY
To talk about Cloud Computing, it might help to first define
it. The difficulty is there is a vivid
debate occurring about just that subject.
The notion of a cloud (as a technical concept) first
appeared in communications. Every
diagram of networks, particularly multiple networks, was topped with a
“cloud.” The idea was that the cloud was
beyond specific definition, but that it included everything that the network
might need, and changed over time as technology and applications evolved.
It was natural for the biggest and most pervasive network,
the Internet, to come to be thought of as a cloud.
Cloud Computing takes the idea of processing-at-a-distance
(not a new idea at all) and translates it to the world of the Internet. Cloud Computing has its precedents in time
shared computing, when in order to connect any remote user to the computing
environment, expensive, secure physical connections had to be created. In the Cloud Computing round, we substitute
the Internet (with or without secure connectivity) for those specific physical
connections. This changes the equation
enormously and makes access to the function of the Cloud ubiquitous.
In the beginning it was easy. We had some clouds. They offered access to compute power and
storage at a fee, usually per user per hour.
All the clouds were public, in the sense that anyone with money (or more
realistically, a credit card) and an Internet connection could use Cloud
Computing. But the term Cloud Computing
quickly stretched to include not just the processing offering, but other
services, as well as applications (previously referred to as Software as a
Service or SaaS) that the clouds enable.
Initially, I diagrammed this Cloud environment as a layer
cake, adding first management and later an application integration layer (as
yet largely unrealized).
|
Integration of Applications
|
|
SaaS Software
|
|
Management
|
|
Utility Computing (Infrastructure)
|
But after a number of discussions with Michael Salsburg, a
technical executive and architect at Unisys, I redesigned the “layer cake,” to
more accurately represent the relationships between the various parts of the
Cloud. In this diagram, Management moves
from the layer cake and becomes a vertical layer affecting all of the other
parts.
|
Management
|
Integration of Applications
|
|
SaaS Software
|
|
<------------------------------------------Management
|
|
Utility Computing (Infrastructure)
|
Personally, I felt it was crisper to refer to the
infrastructure as the cloud and the applications as SaaS, or cloud-based
software, or whatever. But it’s clear
that this argument is lost. All of the
things in the diagram above, singly and in combination, are not being referred
to as “cloud” or “cloud computing.”
But nothing appealing remains static and unchanged and Cloud
Computing quickly grew in multiple dimensions.
1.
Early Cloud Computing vendors like Amazon and
Google added additional functions (often for additional fees) such as storage
and development software.
2.
Additional vendors entered the marketplace, such
as Rackspace and OpSource, each with its own focus, based on its previous
experiences, customer demand, and strengths.
3.
Large customers who were interested in the idea
of Cloud Computing, but unwilling to place their information in a shared
environment, wondered if “private clouds” were possible.
It is this interest in attracting enterprise business and
providing additional security and privacy that has caused much of the
controversy over the boundaries and definition of Cloud Computing.
Many Cloud Computing providers will now provide a “private”
cloud for a customer, usually a large enterprise. Until recently, this meant building a new physical
data center around a cloud architecture (virtualized servers, flexibility in
scaling up and down, and the ability to support (and connect) users in many
geographies. But it was entirely
private and “owned” by the customer.
Ownership, of course, is a flexible concept itself. IBM, for example, will build infrastructure,
including private clouds, for customers and charge for them on a lease or
lease-to-own basis as well as selling the new facility to the customer.
Although access to this private cloud would normally be via
a VPN over the Internet, the cloud was invisible and unavailable to anyone
other than its owner and his invitees.
Scalability was limited to the resources of the private cloud and the
customer was making an investment in capital infrastructure (or at least a
commitment to it).
Recently, several vendors (Amazon and OpSource, for example)
have started offering “virtual private clouds” (or “private virtual
clouds”). In this case, the customer
contracts with his cloud vendor to create a private cloud within the vendor’s
larger public cloud. Subject to contract
terms, the customer can choose to scale his private cloud up and down, enjoying
the capital infrastructure investments and larger resources of the vendor,
rather than being limited to his individual organizational initiative. From within the public cloud that enables his
private virtual cloud, the customer could choose to interoperate with other
clouds (or other web-based applications).
And there’s the controversy:
Is a private cloud a cloud at all?
Some claim that since a cloud, by definition, offers the ability to buy compute power (and
other capabilities and services) on demand, without the need to buy
the infrastructure, a private cloud is by definition not a cloud at all. These same debaters usually argue that a
private virtual cloud is a cloud because it is simply a
temporary (although temporary could be as long as the customer likes) piece of
a public cloud, subject to scaling up and down, as any proper cloud computing
offering is intended to do.
It’s this kind of debate that convinces me that Cloud
Computing is still pretty immature. Not
that it doesn’t work; it does. In fact,
lots of customers are happily computing away in the cloud, for purposes as
diverse as additional compute power, development and test environments of any
required configuration, access to software more immediately and with fewer
internal resource requirements than traditional, in-house implemented software,
and the ability to create shared environments, outside the organizational
firewall, where all kinds of workers (employees, contractors, suppliers,
customers, and others) can collaborate.
Rather, we are all still testing the boundaries of exactly
what we’d like Cloud Computing to be.
Vendors want to take advantage of a hot new marketing term; customers
bring their normal expectations to a new technology: they expect it to solve every problem. Cloud Computing is no different than other
technologies. It can offer extraordinary
and growing capabilities that will make it attractive to most business
customers for some things. But no
technology will ever solve every problem and it is when the users recognize the
boundaries of a new technology and routinely use it for its strengths that a
technology begins to mature and be genuinely useful.
