IAAS and PAAS: Getting Closer all the Time
By John Considine
Happy New Year! In this first post of 2011, I’d like to explore one of the primary ways the cloud landscape is evolving. Two of the pillars of cloud computing, Infrastructure as a Service (IAAS) and Platform as a Service (PAAS), are showing some interesting trends as cloud providers adapt to meet the needs of their customers. Over the coming year, we may see these familiar models evolving into something new since the ideal solution for most enterprises is not one approach or the other but some combination of both.
Traditionally these two methods of cloud computing have been quite distinct. Infrastructure as a Service providers like Amazon EC2, Terremark, and Savvis promise to remove the burden of managing physical infrastructure — everything from server installation and support to network and storage infrastructure build-out and management. Platform as a Service offerings like Force.com, Google App engine, and Microsoft’s Azure provide these benefits of IAAS in addition to offloading management of the underlying system and application software. Operating system software, core services, and even high level application building blocks are managed by the provider, freeing users from worrying about things like patching, updates, core application configuration and management.
Many feel that IAAS does not go far enough to eliminate the unnecessary overhead of managing common software components. Significant effort is involved in managing operating system lifecycles and common software components, and the PAAS supporters are pushing for cloud computing to eliminate this wasteful effort.
PAAS also has its downsides, particularly transition costs and vendor lock-in. In order to transition your workloads into PAAS, you have to adopt and design for the specific offering that the PAAS provider has created. This creates the potential for vendor lock-in because your new applications are using the specific services built by your PAAS provider, and you are unlikely to find the same services from another vendor.
We have found that enterprises are using all forms of cloud computing, from SAAS offerings like Salesforce to PAAS offerings from Microsoft to IAAS from Amazon. Within any given enterprise, there are multiple groups, departments, and users that have specific problems and are seeking solutions wherever they can find them. For example, business users in the organization are turning to SAAS to solve their customer management and collaboration, while developers are building new solutions on the PAAS platforms to speed development, and IT ops teams are utilizing the rapid provisioning and scalability of IAAS to get their work done. It is perhaps the pressure to provider broader solutions for these organizations that is leading to some interesting changes in the services offered by IAAS and PAAS cloud providers.
From IAAS providers, we’re seeing a trend to offer more PAAS services. This is apparent in Amazon’s offerings as they add services such as Relational Database Service (RDS) and Elastic MapReduce (Hadoop) to their SimpleDB and Simple Queue Service (SQS). These higher level services extend beyond simple IAAS into the realm of PAAS since they are not plain virtual machines but full services managed by Amazon. They carry all the benefits and disadvantages of PAAS offerings, with the interesting characteristic of being integrated as part of the overall Amazon solutions.
At the same time, we are seeing a new offering from Microsoft on their Azure platform, the VM Role. This soon to be released offering from Microsoft extends the PAAS nature of Azure into the IAAS realm. Now the customer can control all aspects of instances in Azure including all OS configurations and settings. Of course, this carries the downsides of IAAS as well, in that the customer must now completely manage the OS and applications.
These two providers give us insight to how the market is evolving — IAAS and PAAS are converging as vendors in each space adopt characteristics of the other. This convergence makes a lot of sense since enterprises want just as much control as they really need and the ability to offload work to a provider whenever feasible. In cases where they need to control the entire environment, they want low-level control. This is good news for both providers as well as customers because providers can offer value-added services that are “sticky” for their customers, while different groups within a given enterprise can choose the level of service they want, from fully-managed to completely controllable, now perhaps from the same provider.
As IAAS and PAAS morph into a converged model, CloudSwitch provides the position independence and flexibility that allow enterprises to take advantage of this evolving market without having to adapt to each cloud provider. By making the entire application stack (or selected portions of it) completely portable, from low level infrastructure to high level application control, customers can run their applications where it makes sense, with the management capabilities they need. It’s all about giving customers the choices they want, without risk of lock-in.