Welcome to Cloud 101, a course on Cloud Computing. The intent of this course is to provide buyers of Cloud Computing with enough knowledge to purchase the appropriate Cloud Computing service for their business, and deploy their software on it. The class is going to be created on-the-fly in a blog format, so your feedback and comments will help to direct it towards your needs.

- a basic understanding of what is inside a computer system (CPU, memory, disks)

- a basic understanding of computer networking and the Internet

- a basic understanding of how software is developed and deployed

Feel free to skip any lesson in the series, or to jump around as your interests guide you. The first paragraph of each lesson will give you an idea of the topic material it covers.

Ask ten people what they think Cloud Computing is, and you’ll get ten different answers. This makes it exceedingly difficult to talk about it. It also makes it difficult for customers to compare, buy, and pay for it as well as for vendors to sell it. So before we get into the details, it makes sense to agree on at least a general definition of what Cloud Computing is.

Everyone talks about three basic characteristics of Cloud Computing:

1) Abstracted: In the past, if you wanted computing you’d buy or lease a computer. With the hardware (the computer) abstracted into service, you can now buy computing, which is what you wanted in the first place!

2) Scalable or Elastic: Unlike older ways of buying computing – where you had to buy a computer which gave you a fixed amount of computing power – with Cloud you can scale your computing usage up and down as your needs change and avoid having to overbuy to prepare for future needs. In the past, scaling meant changing out hardware, but now it’s as simple as clicking a button on a web page, or even just setting a policy whereby the amount of computing follows the demands for it.

3) Pay-as-you-go or utility billing: Since you can change how much computing you buy over time, it makes sense that it be sold to you like electricity, where you only pay for what you use

There are other characteristics that people often rely on and hold important about Cloud. These characteristics depend a lot on which cloud vendor you choose, and we will discuss them further during the class.

4) Uptime: Cloud is often assumed to be more reliable than simply leasing or buying a server or servers and putting them to work in your business, in part because experienced professionals are providing the service, and in part because the technology used to provide it is more sophisticated than simply one or more computers hooked together.

5) Security: Cloud is often assumed to be more secure or less secure than owning/leasing your own computer – depending on what the expectations are.

6) Cost: Cloud is often assumed to be cheaper than owning/leasing your own computer. Also, TCO (total cost of ownership) of a cloud-based IT strategy is assumed to be lower than having an IT department that owns its own computers.

People often categorize Cloud Computing into three levels of use model (or Cloud Computing services.) All of them satisfy the three characteristics listed above (#1-#3). The chart below shows how the three levels of cloud computing create the service the end user of the application software sees, and how it was provided before the days of Cloud Computing.

1) Infrastructure-as-a-service (IaaS): Cloud Computing replaces bare computer hardware. Customers manage their “cloud computers” from a system administration level, but don’t worry about setting up or maintaining any hardware. The user of IaaS is the person who would have bought the computer before Cloud Computing existed. Examples are companies such as a Amazon, ENKI, GoGrid.

2) Platform-as-a-service (PaaS): Cloud Computing replaces an execution environment for a computer language by providing a system ready to execute the user’s software. The user of PaaS is the programmer. Note that by installing the execution environment into an IaaS service, you produce your own PaaS service. Many choose to go that way because they have more control over the software environment. Examples are companies such as Engine Yard or Google.

3) Software-as-a-Service (SaaS): The user interacts directly with the Cloud-hosted software, and often pays for “seats” or “users” instead of computer time. Examples are NetSuite, SalesForce.com, Google Apps

These models are not exclusive, and can even be combined in many ways: for example, you may use SaaS from a company that purchases IaaS, or you may use software that your own administrators maintain but runs on an IaaS service from a cloud provider. How you use Cloud Computing really depends on where it can make a positive difference.in your value chain (how you do your work or create your product/service.)

This course will focus on Cloud Computing that is sold by the resource-hour, such as a CPU-hour or Memory-hour, Disk-usage-hour, etc. which is typically IaaS.

Another way to look at Cloud Computing is that it is a way to bring the benefits of the latest technological advances and economies of scale that a large corporate datacenter would have, to customers who cannot afford such a large expense. These technological advances include computer virtualization, multi-core servers, I/O virtualization, high capacity and performance storage, and automated management of the datacenter to provide failover and disaster recovery automatically. Large datacenters also enjoy economies of scale by eliminating middlemen and buying in bulk, savings which the average co-location customer simply cannot access.

Cloud Computing is a way to aggregate demand for data center services run under a highly automated virtualization management framework in such a way that users enjoy economies of scale and best-of-breed IT experiences without having to build the infrastructure for themselves.

Last week witnessed some interesting pronouncements at the Structure09 conference, in which the industry's leading lights had the chance to boast about their accomplishments.  One that caught my eye was that LinkedIn "only" uses 350 servers.

I'm shocked to hear that a text-based service, even with millions of users, uses 350 servers, and that someone would think that was a relatively small number.   Working here at ENKI with customers who do similar things to LinkedIn, I'd always assumed it would be 20-30 servers, at most.   I can't help but think that LinkedIn isn't taking advantage of the latest automated virtualization management technologies, or there is some sort of inefficiency in their system.  I'd love 350 servers of their business, but I think I'd be embarrassed to charge them for all those servers.
As an example, I recently participated in planning the conversion of a major email management site to our cloud.  They have about 40 servers running a windows-based infrastructure.  I believe, as do they, that their software implementation is relatively efficient. In our cloud, after removing clustering and failover pairs - no longer necessary due to automated hardware failure recovery that brings the app back online within the same amount of time as a cluster can fail over - they only needed 20.  When we added automated workload management, which scales actual server use to demand, they averaged 4 servers over the course of a day (which is what they actually pay for.)   That big reduction came because their demand was highly variable, so their servers had averaged 5% utilization over the course of the day.  (See graphic at end of this blog entry.)

Overall, they saw a 10x reduction in the number of servers needed to run their site.  That's a cost savings for sure, but also 10x less greenhouse gases, something to be proud of.

Applying a similar method to LinkedIn, they should be at around 35 servers, using technology similar to what we use (whether in their datacenter or ours.)  What am I missing here? 

I was responding to a blog post this morning relating how Matt Mullenweg, The founder of WordPress, recently stated that open source on leased servers was superior to Cloud.  My mind reeled at the assumptions in that statement, but mostly I was reminded how even the leading lights in the industry still don't understand what cloud computing is.   

In the process of replying to the post, I realized that a more general definition of cloud might be what the industry and customers needed in order to promote useful discussion, instead of "my orange is better than your apple" (note the lower case on that!)

In my view Cloud is a way to aggregate demand for data center services run under a highly automated virtualization management framework in such a way that users enjoy economies of scale and best-of-breed IT experiences without having to build it for themselves. 

A few weeks ago, I attended the Cloud Camp at Sun's Java 1 conference.  I learned something that really surprised me about Cloud Computing that I'd like to share with you.

I entered the conference excited about sharing some of my experiences as a Cloud vendor, and hoping that those experiences would benefit others at the Camp.   I'd already decided that I'd like to discuss "how to monetize your application in the cloud" as a topic, since so many of my customers ask me questions about this.

The Camp, being an un-conference, started out with the audience listing their questions and the moderator writing them down.  Then the moderator asked any members of the audience if they had expertise to address the questions.  I did, so I ended up on the panel answering questions.   While there were many questions about Cloud Computing, most seemed to focus on deployment issues such as cost and security.  Eventually we narrowed the questions needing more discussion down to a small subset, and broke up into groups to discuss them.  Since I'd suggested monetizing your app in the cloud as a topic, it got the votes and I ended up moderating a discussion session with that topic.

However, we barely were able to touch the topic.  Instead, almost every one of the 20 or so people in the group seemed to be more concerned about cost than revenue, even though the topic covered both.   Some were upset at the cost of EC2 which they had been using.  Others were confused about what "on-demand" means.  Others didn't understand how they could predict their costs.  And many just wanted to listen, frustrating my desire to have a stimulating conversation.  Almost everyone was confused about exactly what "Cloud" meant.   I ended up teaching a mini-class on how the choices they make end up resulting in the costs they have to pay for Cloud Computing, and what they can do to minimize cost.

This ended up pointing out to me that the Cloud industry has a problem.   If customers don't know what the product is, and they have high expectations of it as a result of lots of advertising and hype and excitment, they are bound for disappointment, and vendors are bound to have a hard time making them happy.  And the people at this Cloud Camp - unlike the geek-focus of prior Camps I've been to - were CTOs, Directors of IT, VPs of Technology and the like.  And they were confused.  If they didn't get it, how would COOs, CFOs, business owners, and entrepreneurs?

This resonated with me, since we founded ENKI on equality of power between us and our customers...  in other words, a win/win paradigm.  You can't have win/win if one side holds all the cards - or all the knowledge.  So for the industry to move forward, we will need to educate the customers.

To this end, I'm creating a "Cloud 101" class that I will be presenting here in this space, and eventually as live sessions at local events in the Bay Area, as well as video segments on this site.   The idea will be to present a vendor-neutral look at what Cloud Computing is, and how to use it successfully.

To continue with the un-conference format, I'd like to solicit suggestions from you on what you'd like to see in the class.  My current straw-man syllabus is:

- The on-demand computing model and the pay-as-you-go billing that results from it.

- Is Cloud Computing less expensive than alternatives?  How does it save you money?

- Reasons to choose Cloud aside from cost considerations.

- How to make money deploying your software to the Cloud?

- How should I write my software to run in the cloud?

- What are the differences between various Cloud technologies?

- Your topic here.

If you have a Google search running on "Cloud Computing" as I do, you will find yourself inundated every day with announcements from companies about new cloud computing services.  From personal experience, I know that creating and operating a viable cloud service is not easy, and requires both technical expertise and experience.  So as a result I think many of these announcements are just hype.  Network World's pundits seem to think so as well, and are giving the hype a sharp tongue-lashing.

While I think the hype around Cloud is bad for customers, it's also bad for vendors because it both creates unrealistic expectations among customers as well as obscuring which vendors can provide the value customers are looking for.  If everyone is "cloud washing" their non-cloud technology, then how can a realistic assessment be be made of it?  And how can customers benefit from the underlying technology revolution that makes Cloud possible?

A great place to start is with a definition of Cloud Computing that eliminates the "cloud-washers".  The Federal Government has reasonably defined Infrastructure-as-a-Service Cloud computing, and their definition can be found in another of my blog articles.

However, this definition doesn't address the main reason that Cloud is so interesting to vendors and customers alike: can it augment or replace the corporate data center while offering pay-as-you-go operations cost savings and reducing headcount and required skillset?

Answering this question still falls into the hands of the buyer, for whom I propose the following questions about potential cloud vendors, though many customers will already have these questions available from building their own infrastructure:

1) Does the cloud achieve the CPU, I/O, and storage performance necessary to run my applications?
2) How does the cloud infrastructure guarantee reliability by design and how much reliability can I expect (as opposed to being promised)?
3) Will I understand the underlying architecture of the cloud well enough to optimize/tune and secure my applications?
4) Is security equivalent to that which I have deployed for my applications?
5) Is the controllability and usage accounting capable of supporting my budgeting process?
6) Do I trust the cloud vendor as much as I'd trust my own IT department?
7) Is there adequate support for any regulatory compliance activities that I wish to undertake?
 
There are probably more, but these seem to be the ones that enterprises ask me over and over.  I think I have good answers to some but not all of them, so I know my work - and that of other cloud vendors - is cut out for me to stand clear of the hype, which seems to float to new heights every day.

I'm a little surprised that I'm writing this, and not some representative of the IT buying community - or perhaps I've missed it with my saved searches.  In any case, I welcome your comments.

Suddenly in the last few weeks, I have received from multiple prospective customers a copy of the Federal Cloud Computing RFI which the government posted a month ago on its purchasing website (the deadline has passed, by the way.)  It's a long, complex set of questions for cloud computing providers that covers Infrastructure as a Service characteristics and standards compliance.   Clearly the people who wrote knew what they were doing, but it's also pretty clear that not every agency and division in the federal government needs to have a "yes" answer to each of the questions: they were probably compiled by summing the potential needs of all agencies to allow the purchasing office to suggest the best solution for each need.

So it's all the more surprising to get the same RFI from commercial customers, since it's highly unlikely that any one of them will have all the same requirements as the sum total of all the federal agencies!  Presenting such a set of requirements to a vendor has two results: it slows the response of the vendor to meeting their customer's true needs, and it may cause the customer to eliminate the best vendor for their needs because they are being compared against others that may have more checkmarks on the list of items in the RFI, but not necessarily a better solution for a particular need.   So, by diverting the discussion away from what customers really need for their businesses to comparing who is the "best" cloud vendor, the true benefits of the cloud may well be denied or delayed to businesses that could start enjoying them immediately. As I often tell prospective Cloud customers, the best vendor is the one with an offering that matches your needs most closely.

Some of you reading this will probably take it as me "whining" that the list of requirements is too difficult to comply with or even reply to.   Fortunately, ENKI was able to reply to the government within just enough time to meet the deadline (boy, were my fingers tired!)  And, I think we have a leading position in meeting many of them, though we certainly haven't answered "yes" in all areas.  What I'm more concerned about is that this federal cloud RFI will increase the cost and completion time for sales, which in turn makes cloud more expensive to the end customer.

As a metaphor, I think we can agree that everyone wants a Ferrari, but most end up with Toyotas.  Why is that?  It's because of pricing.  Pricing is the single most effective way to communicate value from vendors to customers.  And unfortunately, the cloud industry is still too young to have solid pricing for many of the items in the federal cloud RFI, perhaps because as in our case, we see little actual demand for them.  This causes customers to in turn make irrational buying decisions, because the data to make rational ones isn't out there.  Using my analogy, if Ferrari had introduced their cars without a price, people might assume they would cost the same as a Toyota or just a little more, and flock to the showrooms, clogging them with looky-loos who in the end would not buy.

What I *do* think is good about the RFI, since it seems to be circulating widely, is that it will start the discussion on what is important and what is not from the Cloud.   In the meantime, I am going to be spending quite a bit of time explaining to many of my customers why they don't need a Ferrari to take their kids to school!