Cloud Computing and Analytics

At the time of writing, a Google search on the phrase "cloud computing definition" returned more than 3.5 million results. There appear to be as many definitions of cloud computing as there are people excited about it! Some of these definitions are very good. For example, the U.S. National Institute of Standards and Technology (NIST) provides a concise but comprehensive effort at http://csrc.nist.gov/groups/SNS/cloud-computing

This book will not repeat such efforts at defining cloud computing. Instead, we intend this book to be a practical companion to leveraging cloud computing in your IBM Cognos analytics solution. As such, it focuses on the main characteristics of cloud computing with respect to their tangible advantages for you, the cloud practitioner.

On-Demand Infrastructure

On cloud computing platforms, the required IT infrastructure for your applications is provided to you, based on what you actually require. Nearly all clouds now provide compute cycles, networking, storage space, and memory capacity, all on an on-demand basis. Because you can simply release unused resources back into the pool, you do not have to worry about over-purchasing more hardware than you actually need.

In a pure and simple comparison with traditional data centers, this arrangement provides immediate and obvious cost advantages. Underutilization of purchased hardware is a genuine problem. It's what made virtualization such an attractive IT strategy in the early 2000s: replace the physical hardware with virtual hardware so you can allocate virtual machines when you need them and deallocate them when you're done. This strategy is particularly cost-effective for analytical applications that are tied to seasonal behavior, such as a sales application that is used only during the end of a quarter.

Small wonder that the major cloud platforms, including those from Amazon and IBM, are, at their lowest level, Web interfaces wrapped around virtual machines (VMs), storage, and networking. Being able to create and configure VMs through a simple browser interface or through Representational State Transfer (REST) calls is one simple way to think about and approach cloud computing.

This pay-as-you-go, utility-based cost model is, in some ways, the most innovative aspect of cloud computing. You trade away the requirement for up-front capital expenditures (capex) to purchase hardware and software, and instead favor ongoing operational expenditures (opex) based on what you actually use.

This book takes you through the process of leveraging such an infrastructure to create a fully working IBM Cognos Business Intelligence (BI) virtual instance running in the cloud. This instance is then saved as an image, consuming no resources or cost, until you are ready and have a need for a Cognos deployment.

While the steps in this book are based on the IBM Smart Business Development and Test Cloud, they are also applicable with little modification to other cloud infrastructures, such as Amazon's. And of course, the best practices we describe here have general applicability and relevance, no matter how you ultimately deploy your Cognos application.

On-Demand Higher Services

Moving above the so-called Infrastructure-as-a-Service (IaaS) layer to the higher so-called Platform-as-a-Service (PaaS) and Software-as-a-Service (SaaS) layers is where cloud starts to differentiate itself from simple virtualization. The PaaS and SaaS cloud layers bring higher-level services to the table, and things get much more interesting. Rather than thinking in terms of machines, networking, bandwidth, and storage space, imagine services related to the provisioning of complex topologies with defined quality-of-service constraints, analytical and reporting services, and Hadoop-style big-calculation jobs.

While nascent, there is a tremendous amount of growth in the cloud computing ecosystem around these higher-level cloud services. For example, both Amazon and Yahoo! offer platforms that can execute Hadoop applications in their cloud infrastructures.

There is a good and practical reason why higher-level services are interesting: they are more cost-efficient. For example, running a Hadoop job using Amazon's Elastic MapReduce, which leverages Amazon's Elastic Compute (EC2) and Simple Storage Service (S3) under the covers, costs less than directly using EC2 or S3 yourself. That's because not only do you avoid having to install and configure the software, but Amazon can optimize and manage the entire infrastructural stack much more efficiently.

The end result is that as we move up the cloud stack and focus more on higher-level services that provide targeted solutions and workloads, we are able to build more for less.

Resource Pooling and Rapid Elasticity

The distressing amount of hardware underutilization in traditional data centers that we noted previously remains the main reason why virtualization and cloud has been on the IT agenda for the past few years and will continue to be in the years to come. Being able to more closely match capacity and cost with demand is the cost justification we all need.

This ability to pool and share resources to match demand clearly requires rapid elasticity. The amount of storage available (and being paid for) should always be slightly ahead of the growth of your database. Any new virtual machines required to handle added load should be recruited and connected to the system in minutes or hours, not days or weeks.

This book provides techniques and best practices to scale up or down a Cognos BI system, dynamically recruiting additional virtual machines and connecting them or disconnecting them as required.

Flexible Deployment Models

When most people first hear of cloud computing, they think of the public cloud — an IT infrastructure that is delivered externally. While a public cloud is appropriate for many scenarios, there are many other cases in which data cannot leave the enterprise boundaries. Fortunately, in addition to public clouds, you can deploy your solution to private clouds, where the cloud infrastructure is erected within the enterprise firewalls and managed by the enterprise IT department itself, or even to hybrid clouds, which are a combination of public and private clouds, with systems on both sides and a secure connection between them.

No single model will work in all cases. Data security and sensitivity, bandwidth and latency, and even legal and regulatory requirements all need to factor into the deployment and topology of this solution. Fortunately, cloud computing offers a flexible allocation of resources and the ability to loosely couple systems together with standard Internet protocols, letting you design the right solution for the requirements on hand.

This book provides some...