jump to navigation

IaaS vs PaaS June 10, 2010

Posted by spdguru in Infrastructure as a Service, Platform as a Service, Software as a Service.
add a comment

I understand that this particular discussion has been brought up many times among the technorati. It is however important in shaping future Cloud centric applications. One of the parts of the Cloud Computing definition by NIST is the SPI (SaaS/PaaS/IaaS) model, which is definitely a good start. It helps us understand the distinction between various Cloud offers. In very broad terms, when a provider allows a user to control the operating system of the hardware the offering is Infrastructure as a Service. An example of this is Amazon’s EC2 offering. When a user (application developer) has no control on the Operating System (OS) but does have control on the complete application software, and uses application development services like those from the App Engine from Google, the cloud offering is a Platform as a Service. Finally, if the end user has no control on the Application software, except some configuration, like in the case of Salesforce.com, the offering is classified as Software as a Service (SaaS).

The line between IaaS and PaaS is very gray. Amazon’s offerings are deemed to be IaaS and this is probably true in the case of EC2. But Amazon also offers database services (SimpleDB & RDS) and a billing service (DevPay)which are clearly beyond IaaS. Even the S3 storage service provides read write capabilities, which are generally the domain of an OS. The PaaS definition should be broader than the ability to develop an application and physically run on the platform. The Apple App Store surely is a market place, which allows developers to showcase applications and monetize them. The App Store should be a platform too, since it plays a very important role in the lifecycle of an iPhone or iPad application. It is not a stretch to envision the emergence of specialized platforms targeted for industry specific applications. For instance, Authorize.net provides a set of services for credit card payments accessed over the Internet or “in the Cloud”.


Cloud Computing Essentials June 3, 2010

Posted by spdguru in Cloud Computing Essentials, Infrastructure as a Service, Platform as a Service, Software as a Service.
add a comment

Starting with Cloud Computing Essentials – why?

If you have been reading, writing or dabbling in any sense with the topic of cloud computing, you would agree that if you ask 10 people about cloud, you would generally get 15 different definitions. So I thought, let’s start with the ‘cloud essentials’ comprising a cloud ‘definition’ and then figure out cloud ‘components’. This will give us a good foundation to start with.

BTW, elasticity here refers to the cost-efficiency when scaling up or down. As one could imagine, being cost-efficient whilst scaling down can be hard to achieve but feasible with the right cloud-based IT services offers.

Start with a Cloud Computing Definition

Now we have a definition, let’s visit cloud ‘components’. Basically, I have developed two compartments for the cloud-based IT resources and services– how businesses will utilize and how providers will deliver.

NIST has done a great job of providing definitions and a presentation here.

Cloud Service Providers and Cloud Users Components

I often refer to the above chart as a ‘swim lane model’ where the top 2 swim lanes refer to how a business with first justify why it will utilize cloud-based IT services and resources and then systematically figure out how the business will deploy such IT functionality.

For example, the 5 key characteristics of the cloud usage (NIST) are important for the business to understand from both technology and economic perspective. I think self-provisioning through a portal will definitely be the ‘make or break’ factor for the acceptance of any cloud-based offer as it really defines the ‘quality of experience’ for the cloud user. Better the quality of experience, more likely the users will flock to the cloud provider. Cost will always be a part of the success as well but then cost is always an important factor, regardless of the economy at the peak or in the trough.

The business deployment model for the cloud-based services and resources really speaks to the governance (public or private provisioning), lock and keys to the physical location of hardware running ‘the cloud’ (internal/external) and who deploys the security policy (bounded/unbounded). A great example is collaboration provided by Cisco Telepresence and Cisco WebEx solutions, actually.

Cisco TelePresence is Private, Internal and Bounded deployment of IT services and resources providing collaboration. This is because a collaboration session requires provisioning done by authorized parties, access to telepresence room with a physical badge or such and pertinent IP traffic is secured by known security measures such as firewall, IPS, DDoS Mitigation, encrypted tunnels, and such.

In comparison, Cisco WebEx provides collaboration via Public, External and Unbounded deployment of IT services and resources. Cisco WebEx has the multi-tenant asset of MediaTone Network. As a result, any party with an active service account and correct credentials can provision a meeting on the IT resources and services physically residing in the MediaTone Network and be secure as each meeting ‘join’ is secured with SSL encryption with the closest MediaTone Data Center.

By the way, Private, Internal and Bounded define Virtual Private Data Center (VPDC) here.

A couple of points to clarify that changing the deployment element to External means that the VDC is hosted and managed on the SP premises and Unbounded means it is secured according to the business security policy by the provider.

Now from the cloud delivery point of view, there are 3 swim lanes.

The first one is Delivery Models. The delivery model is known as the SPI (Software, Platform and Infrastructure) as a Service model. Currently, IaaS model is being build out by Savvis, Terremark, Rackspace and others.

The second one is Payment plans. Self-explanatory Pay-per-drink and All-you-can-eat models are generally in place. Pay-per-drink model is attractive but the well known $0.10 cent per VM per hour does not buy you a production class server to run your critical application, say a transactional database. However, according to the InformationWeek in the report ‘The Pubic Cloud: Infrastructure as a Service”, generally $499 per month for 1xCore CPU, 4 GBs of memory and 32 GBs seems to be a commonly found tariff.

Finally, the integrated management that is critical to get right for the quality of user experience that we talked about in the beginning.  Having a cross domain manager (physical and virtual resources), OFAB (Operational Fulfillment, Assurance and Billing) and service modeling (within a customer cloud pod or across a multi-tenant cloud cluster) are very important. On a future date, I plan to discuss this more in detail as I think this is the ‘secret sauce’ that a cloud provider can create to ‘make or break’ their cloud offer.

The next topics I plan to discuss are cloud security, service orchestration and top 5 IaaS offers.