When should you own or rent your own cloud (vs. simply sharing one with others)?

We all use cloud computing on one scale or another (from marketing via Twitter to owning our own corporate server clusters). This posting outlines a decision model to help you decide whether you should own, rent or share a cloud to host each of your computing services.

Wait a Minute: I don’t use Cloud Computing

Cloud Computing is simply the process in which one group provides computing services to a group of users in a “turnkey” manner, i.e., in a manner where the end user does not have to worry about how the services are managed, provided and maintained and can get them “on-demand.” In its most basic form, any time you are repeatedly using computing services (i.e., using them without requiring a supporting IT project) that are not hosted on your workstation, you are effectively accessing a cloud of services (vs. twenty years ago when I had to host most of the applications I was using on my own workstation.)

cloudcomputing

Even when we use a basic email service provided by the guys down-stairs we are using a local cloud of our very own. When we use more universal services like Twitter, we are using a truly global cloud we share with many others. In essence, today everyone is using one form of a cloud or another. The real decisions are–

  1. Who provides the cloud? and
  2. Do I pay for this?

Three ways to obtain Cloud Computing services

There are three basic ways you can obtain Cloud Computing services for your enterprise:

  • Own Your Own Cloud. Your enterprise owns the hardware and the software hosted on it. You manage it yourself, either with staff or direct contractors. You control what is installed in this cloud, when it is installed and how it is managed. Example: your corporate Microsoft Exchange Server
  • Rent a Cloud. Your enterprise licenses services from a Software-as-a-Service (SaaS) provider. This SaaS provider manages these services on your behalf but provides a level of dedication through a managed Service Level Agreement (SLA). Examples: Sales force workflow management from Salesforce.com and Web Analytics from Omniture
  • Share Your Cloud with Others. You sit on an open, freeware service with the rest of the world. You have no dedicated support or SLAs other than what everyone else gets. Examples: IdeaScale for crowd-sourcing or WordPress for blogging (when it is provided for free via your hosting provider)

Deciding how to obtain Cloud Computing services for your enterprise

This is not an “all or nothing” decision, i.e., it is unlikely you will use one model to obtain Cloud Computing for everything you need. Instead, you will need to examine each of the computing services you need and determine whether to own, rent or share each. I recommend looking at three attributes of each service to make this decision:

  1. Importance of this service to your enterprise
  2. Level of Annual Investment you have budgeted for this service
  3. Target of Opportunity the service poses to attack by hackers or competitors

1. Assessing the IMPORTANCE of each service

Not all services are equal. Some are critically important; others are “nice to have.” I have found it useful to organize your services into three priority-based buckets using a simple litmus test for each:

importance

Note: You will find this litmus test dovetails well with compliance with regulations such as SOX (Section 404) and SAS-70.

2. Determining your LEVEL OF ANNUAL INVESTMENT for each

Cloud Computing operates of the Law of Large Numbers, essentially obtaining operational efficiencies of scale through size and passing these along to customers. As such, the size of your enterprise’s information technology and services (ITS) budget for each service (not the raw size of your enterprise) is critical to determining what type of cloud you use. If you have a small ITS budget you may gain more value from renting or sharing a cloud than building and managing it yourself. On the other hand, if you are huge enterprise it may cost- and risk-prohibitive to ask an external party to scale to support your needs. I organize Level of Investment into four categories based on the combined total budget (capital and expense for technology, services and staff) for each service:

  1. Utility (>$10m USD per year). The service is treated as a utility with its own teams, hardware, etc. You are making it core to your business
  2. Infrastructure ($1m to $10m USD per year). The service represents a major infrastructure item for your enterprise. Its investment is likely made up of multiple line items (for hardware, software, consulting and staff) in your annual budget
  3. Budget Line Item ($100k to $1m USD per year). The service has its own line item in your budget (an likely has performance targets to justify the inclusion). However, you are not spending enough to build and maintain a full infrastructure with its own support team.
  4. Expense Item (<$100k USD per year). The service does not likely reach the threshold for capital planning purposes. You do not have (or cannot afford) full-time staff dedicated to managing the service. This could be a service used by a SOHO or for a trial, experiment or campaign)

3. Assessing the TARGET OF OPPORTUNITY to hackers and competitors

Simply knowing the importance and size of your service is not enough to make your cloud decision. You also need to consider the visibility of this service as a target of opportunity. This is important, because visibility directly correlates to risk of attack from hackers. You may have a small IT budget but be a household name who attracts hackers like flies to honey. Here is my simply litmus test to determine you Visibility Risk Level:

  1. Are there competitors (if you are a public sector agency read this as “other governments”) ACTIVELY trying to obtain the information managed by this service? If so, it is a High Visibility Target
  2. If Service is not a High Visibility Target, would a 14-year-old get “bragging rights” if he or she hacked this service? (I recommend asking your children, nephews or nieces the answer to this question.) If so, it is a Moderate Visibility Target
  3. Everything else is a Low Visibility Target

Now we have enough data for a recommendation

Sit down and list all the services you have that require computing support. For each, specify the following

  1. What is its priority: Mission Critical, Operational or Non-Essential?
  2. What is your Annual Level of Investment
  3. What is thee Visibility Threat Level: High, Moderate or Low?

Then look up each on the table below for a recommendation as to how to source your cloud:

matrix

General notes, observations and disclaimers:

  • These recommendations skew towards not owning anything for which you are not willing to make a major (i.e., utility-scale) investment. The rationale is two-fold. First, if IT and computing are not your core business, you should not tie up investment capital here (save that for the true reason for your existence.) Second, if you are making small investments you cannot benefit from the law of large numbers (as such, it will be very costly to maintain 24×7, high-availability operations).
  • These recommendations steer you away from using shareware for mission-critical services (with the exceptions of campaigns, trials or  small operations). The rationale is simple: you get what you pay for. (Do you really want to place the reliance for the core of large business operations on a freeware application with no SLA?)
  • These recommendations encourage use of shareware for trials, campaigns and experiences–regardless of service importance. Why? This makes trial and error inexpensive. Once you find something that works, you will likely turn up the investment, causing you to upgrade to Rented or Owned services.

Finally, treat this matrix as set of guidelines only. When in doubt, apply common sense and business analysis.

6 thoughts on “When should you own or rent your own cloud (vs. simply sharing one with others)?”

  1. @Joe: To answer your last question, this Blog is a SOHO operation for me (not my core business). Based on the matrix above, I use WordPress to publish (free via GoDaddy co-registration) and Twitter to market my posts.

  2. Good article. One thing to note is that even if there a rent decision made for SaaS computing, there is a point of inflection after which it is cheaper to pull the service in-house. If you choose a rent decision, insure that you have computed the inflection point. When that is reached, either you can re-negotiate your contract with the vendor or simply pull the service in-house.

  3. @JP. You point is “spot on.” It you “rent” a service and it grows large enough that you want to make it a core part of your enterprise (as HGTV did with their Rate-My-Space community) you need to ensure your SaaS provider enables export of: fact data (e.g., accounts) and metadata (e.g., rules), static content (e.g., instructions, articles). Otherwise it will be impossible to leave the rented cloud (you will be in what we used to call a “data jail” at AOL)

  4. Good article. One thing to note is that even if there a rent decision made for SaaS computing, there is a point of inflection after which it is cheaper to pull the service in-house. If you choose a rent decision, insure that you have computed the inflection point. When that is reached, either you can re-negotiate your contract with the vendor or simply pull the service in-house.