For many enterprises, their first experience with cloud computing is using infrastructure as a service (IaaS). IaaS allows enterprises to move many existing applications into a cloud environment “as is,” benefiting from rapid, on-demand capacity and moving from owning and managing the physical infrastructure to “renting” the capacity as it’s needed. 

IaaS’s deployment centers on virtual machines (VMs). Typically this includes several virtual servers with an operating system (e.g. Windows or Linux). The client normally is responsible for installing the application software, setting up the database environment and architecting the hardware and software to support redundancy and high availability. The client also is responsible for maintenance activities, such as patching the operating system and applications and overseeing antivirus, network security and backup systems and processes. The IaaS service provider is typically not aware of the applications running within the environment and only is responsible for keeping the infrastructure highly available; application-level monitoring and support is left to the client.

Platform as a Service (PaaS) represents a paradigm shift from today’s application development practices.   PaaS-developed applications use tools that are specifically designed to take full advantage of cloud computing infrastructure, which results in increased benefits over IaaS; however, it also imposes more constraints on the hosted applications. (Reference my previous post, Defining PaaS: A flexible solution to meet your needs.)

PaaS’s deployment is centered on the application. The cloud provider manages up to the application level and therefore performs many of the routine tasks the client manages in the IaaS scenario described above. This allows developers to focus on enabling business functionality without having to worry about the underlying infrastructure and to take advantage of unique PaaS platform capabilities, such as automatically scaling the application to handle increased demand and then contracting the service when that demand is no longer there.

Since PaaS abstracts the VM from the client, it does create some limitations; in PaaS, the client must accept the features and functionality made available, which may limit the level of customization that can be done to tailor an application to meet a client’s specific requirement.

For those reasons, IaaS works well with most existing applications, while PaaS works best when developing a new application. Here are some examples:

• Bursting – Some applications, such as e-commerce, experience a peak load only during periodic times throughout the year. Provisioning for that peak load means that the infrastructure will be underutilized the rest of the year.
• High performance computing (HPC) – Some organizations require to crunch a large amount of data only at periodic times. Efficient data crunch using HPC can require many servers for a relatively short period of time. Provisioning for those periodic loads could mean buying up to thousands of servers that are used only a few days a month.
• Mobile / social hub – Many businesses foray into social media and mobile technology for the first time with little knowledge of the possible outcome. Provisioning for the best-case scenario can be very expensive.

Those cloud scenarios pose challenges to on-premise infrastructure. Applications running in an IaaS environment can be changed rapidly; however, it requires human intervention.  With PaaS, those scenarios are done automatically without any human intervention.

Microsoft Windows Azure, one of the leading PaaS providers, has an interesting case study with the mobile operator T-Mobile USA: T-Mobile USA: Mobile Operator Speeds Time-to-Market for Innovative Social Networking Solution. In the case study, they outline how they developed a mobile application to simplify family communication and deployed it to the cloud in just six weeks. Obviously this would have been impossible with traditional server provisioning, especially for the unknown load. PaaS gave them the flexibility with a low-entry cost as their new service gathered popularity.

Although PaaS does lower maintenance and monitoring costs, its real power lies in its ability to enable scenarios impossible through traditional computing while meeting aggressive time-to-market objectives.