Design of an enablement process for on demand applications.

Independent software vendors (ISVs) are seeing erosion of revenues because of competition and pressures from enterprise customers demanding applications that are cheaper, less complex, and easier to install. These ISVs must seek new ways to penetrate new markets and grow their customer base. Today's economic and business conditions are requiring that ISVs begin to offer their products as utilities. As utilities, they would be application-level services that are sold on a "pay-as-you-go" basis. To emphasize the on demand characteristics of these services, we call them "on demand services" or "ODSs" in this paper. Unfortunately, ISVs who traditionally sell their applications as licensed products are inundated by information about utility computing and the on demand "hype." They all ask the same question: How do I turn my application into an on demand service?

In this paper we describe the design of a process called the Application Enablement Program (AEP) for on demand services (ODSs), which are software utilities, and introduce ODS architectural concepts. On demand service is synonymous with on demand application. The AEP represents a streamlined, cost-efficient, and globally deployable process. Our design is based on our experience creating such a process for hosting licensed software applications in the IBM Universal Server Farm hosting facility, Version 5 (USF V5). The AEP for ODSs is expected to play the crucial role of transforming ISV applications into ODSs, identifying the tasks to enable the transformation, helping them make the changes to run in the on demand infrastructure and the ODS framework, and finally checking for ODs compliance with the utility platform standards and guidelines.

Background

Before we proceed with describing the AEP, we define some terms and concepts used in the process such as hosting models, utility computing, on demand infrastructure, and the Universal Management Infrastructure to give the reader an idea of where ODSs fit.

Nonutility hosting. Traditional licensed software applications were typically purchased by enterprise customers and installed on their own premises. As software applications became more complex to manage, enterprise customers began moving these applications out of their premises and hired hosting companies such as IBM to host and manage these applications for them.

The majority of currently hosted applications fall into three (nonutility) hosting models. (1) Utility computing is discussed in the next major subsection.

Collocation. The application provider (the customer) collocates the machines running the application with a network bandwidth provider. The application provider is responsible for managing the application server machines remotely or on site. The provider integrates the offering into the existing infrastructure. The network bandwidth provider typically provides the application basic services such as electric power (as an uninterruptible power supply), network connectivity, and bandwidth (power, ping, and pipe).

Dedicated. An infrastructure service provider (InSP) supplies dedicated machines to run a fixed set of applications. The machines may be purchased outright of leased by the application provider. The InSP manages both the content and infrastructure. The infrastructure includes servers, storage, and network connectivity. The dedicated model is a simplified description of USF VS.

Both models typically serve one enterprise customer per single instance of the installation. They require application providers to procure, install, and deploy a new instance of the installation for each new company buying the service. The application usage is also predictable and linear. Both models have severe implementation and cost challenges when demand slows or grows rapidly in an unpredictable fashion.

Enterprise customers pay to maintain their respective resources. They can typically use as many resources as they need, limited only by what their infrastructure capacity permits.

Shared hosting or application model. The InSP provides the bandwidth, servers, storage, and software services for supporting a shared application. The application is shared among multiple enterprise customers with multiple users. The servers are typically owned of rented by the application providers.

This model leverages shared services across multiple customers and applications. The model has a lower incremental cost for each additional customer joining the service. Because the application is shared, the application provider has to ensure that the application scales well when additional customers are added.

Enterprise customers for these shared applications are typically billed on a monthly basis. The use of the services is not metered for usage, leaving the users to consume as much as they want for a fixed monthly fee. Service level agreements (SLAs) may exist, but limits on usage could be difficult to enforce.

Several factors are influencing the move away from these models. Server utilization, application usage, administration complexity, return on investment, total cost of ownership, and economic uncertainty are some reasons that are driving application vendors and InSPs to rethink the overall strategy for business hosting. This is where utility computing can help.

Utility computing and utility hosting. Utility computing in its simplest form is information technology (IT) presented as a utility. It is based on the model of conventional utilities such as telephone service and electricity. The enterprise taps into this utility for business and infrastructure services. The services are physically hosted in utility data centers that can be located either inside or outside the enterprise, of both. Like a traditional utility, the service is pay-as-you-go, charging only for use of the service. It is reliable and caters to the on demand nature of a utility.

Utility computing promises the following benefits: (2)

* Simplify IT by reducing complexity

* Turn IT from a fixed to a variable cost

* Reduce cost or operating expense

Current technological advances have turned the concept of IT delivered as a utility into reality. Some examples:

* Reduced cost of bandwidth enables the creation of new data services and high-speed network delivery of a variety of services to a broader range of customers.

* Distributed content and application architecture deployments shift delivery to the edge of the network.

* Server and storage virtualization enables new levels of shared infrastructures with the potential of reducing customer costs.

In utility-based hosting, application providers pay for the "IT infrastructure power" that they need to run their applications. The application providers pay only for what they use. The utility-hosting provider meters application usage metrics such as bandwidth, storage, and CPI use and bills the application owner accordingly. The utility-hosting provider owns all components of the infrastructure, including servers, storage, and network. This situation is in contrast to nonutility-hosting centers where application providers buy or lease IT infrastructure components such as servers and storage to provide themselves with IT infrastructure power. In a nonutility hosting model, the application provider is more like a consumer buying his or her own power generator instead of signing up with an electric utility provider.

On demand infrastructure defined. An on demand infrastructure seeks to fulfill the IT functionality of today's enterprise business in a utility-like fashion. The on demand infrastructure delivers standardized processes, application function, and infrastructure over the network as a service.

The on demand infrastructure shares the following properties in common with utilities:

* Sharable--capable of serving many customers

* Standardized--requires (allows) little customization

* Flexible and scalable--use what you need and pay as you go

>From a technical perspective, an on demand infrastructure has the following characteristics:

* Integrated--allows the integration of enterprise and legacy applications that transcends vertical industries through the use of open standards technology such as Web services. (3)

* Open--uses open specifications and standards to enable ease of integration of enterprise resources and applications.

* Virtual--uses server consolidation and capacity on demand technology to increase utilization of hardware resources.

* Autonomic--alleviates the need for skilled technical human resources to manage the complexity brought about by the rapid advance of technology. This characteristic also includes responding to customer needs for instant provisioning of resources.

The UMI plat-form. IBM Global Services (IGS) implements an on demand infrastructure by taking advantage of and integrating existing IBM hosting centers with the Universal Management Infrastructure (UMI). UMI is the infrastructure of tools and software that provides services to manage, meter, and provision resources at the hosting infrastructure level. The initial release of UMI concentrates on infrastructure provisioning and managing the provisioned resources. The current version of UMI consists of the following functions: auto-provisioning, monitoring, reporting, metering, billing, and SLA.