HPC Basics Series
Public Cloud computing is a metaphor for a network that allows ubiquitous access to a pool of resources that are consumed and then released. Pricing is based on time and volume resource is usage. There are several different cloud paradigms such as Platform-as-a-Service (PaaS), Infrastructure-as-a-Service (IaaS), Software-as-a-Service (SaaS), and several other variations on the theme of ubiquitous pool access.
What is common among these paradigms, in all cases, the data is housed in a universally accessible location and the computing resources are attached to it rather than the old paradigm of moving your data from machine to machine. In short, the first benefit of cloud computing is being able to be data-centric in terms of file system storage. Now that data is centralized, and compute resources can be attached to the data, this means that compute resources can be scheduled and shared. From a business and cost perspective, this enables an enterprise to consume only what it needs rather than have to pay for, or purchase, an entire machine to do infrequent jobs. Put another way, if you were going to move from one city to another, you wouldn’t buy a moving van, you would simply hire a moving company and rent the van for the duration of your job, the van would then be released back to do other jobs. Cloud computing works the same way, the customer has the option to rent the processing power needed rather than purchasing the machine and paying all the costs for upkeep and maintenance.
In the PaaS paradigm, this notion of renting machine time takes a step further in that the user can now rent commercial applications by the hour rather than paying a yearly licensing fee. The question of “what is public cloud computing?” continues to have an evolving answer, but at its core, the notion of flexible resources are available for rental.
Many would say that Amazon started public cloud computing, but that really isn’t true, a precursor to the cloud did exist back in the 1960’s with mainframes which were started by IBM. A series of lawsuits challenging this business model in the early 1970’s shelved the then practice of renting time on mainframes and required IBM to sell the machines to customers. Thankfully, times and business climates have changed, now even small companies can have access to the massive computing power they need to solve their business challenges.