The greatest challenge for software application companies today is accelerating new license growth while maintaining a reasonable level of expense for sales and R&D. Many have hailed Software as a Service (SaaS) as the best solution to the license growth challenge — after all, SaaS can enable new license sales to a set of customers that historically didn't buy software because they couldn't afford the cost or complexity of the technology.

But SaaS requires a significant investment in technology infrastructure and a change in business model that many application companies can't reasonably undertake. Nor is SaaS necessarily the best option for some customer situations and application profiles. Luckily, a new model is emerging that provides customers with all of the value of SaaS without forcing the application provider to dramatically alter their technology or their business model.

This model is the virtual appliance. Virtual appliances:

• Dramatically reduce the complexity and expense of application installation, maintenance and administration
• Provide application companies with the best features of SaaS and the traditional software model without the associated drawbacks
• Deliver the value of the application without the hassles of technology administration
• Give the software provider a product that is
• easier to sell to a broader market without significant changes to the technology or the business model

In this post, we will further define the virtual appliance opportunity as an alternative approach for delivering SaaS value. We will provide background that establishes the viability of the virtual appliance opportunity, and we will discuss considerations for a software application company to choose virtual appliances as a product format. We will also discuss the challenges that remain in the market for virtual appliances to become broadly accepted as the preferred format for application delivery. Finally, we will argue that virtual appliances represent a watershed opportunity for application companies to fundamentally improve the economics of their business.

The Limitations of SaaS
If Salesforce.com can demonstrate so much success, why have so many other application companies failed in their attempts to deliver SaaS versions of their applications? The reasons for this lack of SaaS success by traditional software vendors are significant and multi-faceted. Historically, the technical hurdles to an effective, multi-tenant SaaS architecture have been considerable (these hurdles are dropping due to virtualization, but more on that later). Re-architecting the application to efficiently deliver it to thousands of customers via the Internet is an enormous undertaking.

Further, software application companies are not generally well-suited to becoming service companies. They tend to have a culture dominated by engineering innovation and license sales execution. The skills and behaviors associated with these competencies do not typically translate well to a customer service culture. Finally, a business model built upon large, upfront license sales to enterprise customers is not easy to transform into a business model that sells subscription services to a wide swath of the mid-market.

Beyond the challenges the vendor faces in migrating from a traditional application distribution model to a SaaS distribution model, there are also significant challenges the customer faces in adopting SaaS for many types of workloads. The primary application value of Salesforce.com is Customer Relationship Management (CRM), which may be the single best application to be delivered via a SaaS model for the following reasons:

• Sales activities are pretty uniform across a wide variety of companies and industries
• The data model and the system transactions are fairly simple
• The users of the system are often widely distributed.

It is a perfect application for SaaS delivery. Other business applications that fit this mold include project management and product life cycle management (PLM).

Many applications, however, are not conducive to delivery using a SaaS model. Business Intelligence (BI), for example, is difficult to translate from the traditional approach to SaaS. The data schemes and transactions are often very complex and the customer requirements for configuration and use are often widely divergent. The workload requires intensive processing, so it is far less attractive for the application provider to assume the burden for the needed high-CPU infrastructure. Finally, consider bandwidth cost and network performance for the large amount of data transfer required for BI applications.

Similarly, Enterprise Resource Planning (ERP) applications do not convert well to SaaS. Like BI, the customer workloads and configurations can be widely divergent. The database schemes and the queries required by the application are often so complex that attempting to host multiple customers on the same database instance (the Salesforce approach to multi-tenancy) leads to terrible performance and significant compromises in functionality across the customer base.

The Virtual Appliance Opportunity
Luckily, the virtual appliance application distribution model delivers all the value of the SaaS model with none of the weaknesses. Furthermore, the market forces that are driving consideration of virtual appliances as a viable model are strong enough to overcome the inertia of the status quo. First, let's define what we mean by the term virtual appliance.

A virtual appliance is an application combined with just enough operating system (JeOS or "juice") and all necessary middleware, with all technology components pre-configured to optimize application performance, delivered to the customer as a virtual machine file ready to run atop a hypervisor. JeOS is not a generic, one-size-fits-all operating system, nor a single product. Rather, it refers to a customized operating system that precisely fits the needs of a particular application. The application's JeOS requirements can be determined manually, or with an analytical tool, such as rPath's rBuilder.

The business logic configuration and all customer-specific inputs to the application are typically entered via a web browser interface - eliminating the need for complex, command line interactions. The application vendor assumes complete responsibility for the life cycle duties of software testing and maintenance for all components in the virtual appliance. The customer receives the value of the application with none of the burden for technology installation, integration, testing and maintenance. The system administration burden is absolutely minimal.

The opportunity for virtual appliances is rapidly maturing because all the elements required for this new model are becoming mainstream. For example, the perfect operating system for the JeOS component of a virtual appliance is Linux. It can be tailored to the unique requirements of an application, and it does not require an onerous contract with a proprietary operating system provider. The availability and flexibility of Linux make it an important driver of the virtual appliance opportunity. rPath, for example, has a Linux service that is easily customized to be JeOS for an appliance.

Another critical element is a high performing hypervisor. A hypervisor is a small operating system that assumes the role of managing all of the hardware infrastructure associated with application delivery. With a hypervisor as an intermediary, applications are no longer constrained to a particular operating system. The customer gains flexibility in its hardware choices, and the application vendor can select and maintain the operating system to suit the application's needs without dictating hardware and infrastructure requirements to the customer.

VMware has certainly demonstrated that the market for high performing hypervisor technology and the associated virtual infrastructure management functions is robust and growing rapidly. With its IPO, VMware is now the fifth most valuable software company in the world. With that type of endorsement from the capital markets, it is inevitable that the legacy performance limitations associated with virtualization will quickly disappear through VMware's investment in research and development.

Furthermore, Intel and Cisco, among others, are betting big on virtualization and VMware. They both invested hundreds of millions of dollars into VMware prior to the IPO to increase the performance of virtual infrastructure on their respective CPU and network offerings. In the very near future, all industry standard servers will come pre-built to support high performing virtualization at native CPU speeds with networking utilities available that enable quality of service and coordination among multiple virtual appliances.

VMware is not alone in the hypervisor space. Both Microsoft and now Citrix, with its acquisition of XenSource, are committed to providing high performing virtualization technology to the market. Both companies will also benefit from the investments being made by Intel, AMD and Cisco in the performance enhancement of virtual infrastructure.

Infrastructure providers will also appear, ready to provide hosting services for virtual appliances such that application delivery can be a variable cost instead of a fixed capital cost. Amazon has already entered this space with its Elastic Compute Cloud (EC2) offering. Without dictating any system software requirements other than a virtual appliance format, EC2 enables applications to run on Amazon's infrastructure for just ten cents per CPU hour. Some technical difficulties remain for the EC2 service to move from a beta category to production status (examples include static IP addresses and persistence of data in the event of a virtual appliance failure), but these challenges are imminently solvable.

The final ingredient for virtual appliance success is customer demand for applications that are easier to administer and maintain. It seems obvious that customers would want applications without the hassle, but the market has built up significant capabilities and infrastructure to cope with the legacy approach to application delivery. There are armies of administrators, consultants and outsourcing companies whose very livelihood depends upon application complexity. Fortunately, companies like Salesforce.com are demonstrating there is a robust and growing market for applications that do not require legions of administrators, consultants and outsourcing companies to maintain them. The rising popularity of SaaS is the final critical element in creating the opportunity for virtual appliances.

Conclusion
If virtual appliances become the de facto distribution format for applications, the economics of the software application industry will be greatly improved. Application companies can deliver their applications to the broadest possible market without any of the traditional R&D and customer service overhead and headaches associated with multi-platform development, testing and integration. Top line growth will improve dramatically because many more customers will be able to pay for the application license when most of the expenses associated with application maintenance and administration disappear. The bottom line margin will improve because 40 - 50% of development and customer service expense will disappear when multi-platform delivery requirements disappear.

For those applications that are well suited for delivery via the Internet as a SaaS offering, virtual appliances, along with high performing hypervisor technology, enable SaaS delivery in a multi-tenant model without creating a special architecture. If the customer wants to migrate the application to their own network, whether for performance, security or other reasons, it becomes a simple procedure.

There are still challenges that remain to reach widespread acceptance of virtual appliances. These include: multiple disk formats for virtual machines; lack of standards for providing services and coordination among teams of virtual appliances; lack of standards for communicating status between virtual appliances and the virtual infrastructure; and, lack of standards around the provisioning, monitoring and management of virtual appliances.

Fortunately, the mid-market does not really care about any of these shortcomings as it relates to their demand for better and simpler application technology. The above concerns are specific to the enterprise market, and the enterprise market already has the best, and most expensive, mechanisms for coping with the complexity of the legacy ISV model.

ISVs should embrace virtual appliances first as a means to better penetrate the mid-market and emerging markets that do not already have significant coping mechanisms for dealing with technology complexity. Simultaneously, ISVs should raise the prices they charge to the enterprise for supporting legacy platforms to encourage these customers to consider migrating to the virtual appliance approach. Ideally, the combination of virtual appliance availability and higher prices for legacy platforms will enable the ISV to gradually retire all support for applications that are not delivered as virtual appliances. The savings in R&D and customer service will then boost the net margin of these businesses by 8 - 12 points - all while growing the top line through market expansion.

Considering the alternative to virtual appliances is not pleasant for the ISVs. It is inevitable that hypervisors are going to become the lowest level of the technology stack - abstracting the hardware and providing infrastructure management services. The ISVs that do not adopt a virtual appliance approach will have to add several more columns to their test and release matrix to deliver market "certified" applications that run not only on the OS and middleware platforms, but also on each of the hypervisor platforms. Instead of being assaulted with more complexity, it is time for the ISVs to fight back by embracing virtualization as a model for changing the economics of the business. If enough ISVs back this approach, the market will certainly accept the new form factor as an approved application distribution format. And the world will be a better place for everyone - vendors and customers alike.

Billy Marshall is CEO of rPath.