Communications of the ACM
Home/Magazine Archive/August 2003 (Vol. 46, No. 8)/Analyzing the Application Asp Concept: Technologies.../Full Text
Communications of the ACM
Analyzing the Application Asp Concept: Technologies, Economies, and Strategies
Technology may have changed dramatically over the past 40 years, but its application appears to have come full circle. Substitute "application service provider" for "time-sharing," and this 1967 definition still works well: "Time-sharing is a communications-oriented method of using computers. It is a technique that permits concurrent utilization of the same installation by two or more persons working at remote devices capable of direct, online access to the data processing equipment" [5].
The application service provider (ASP) concept leverages improvements in network computing technologies to allow desktop computer applications to be hosted remotely from the user. But unlike early time-sharing, the ASP concept supports the modern graphical user interfaces (GUI) users have grown accustomed to. In the ASP computing model, the provider hosts desktop-type applications from a server farm located at its data center. Users from one or more customer organizations access these applications via either a private wide area network (WAN) connection or the Internet. The concept can be applied to nearly any type of application and is commonly used for deploying an array of software ranging from basic office suites, such as Microsoft (MS) Office, to large enterprise resource planning systems, such as SAP. While CPU costs were the dominant economic driver for 1960s-era time-sharing, support costs are the key economic factor for ASPs.
Among the key benefits of the ASP model are economies of scale that allow the ASP to operate a secure, reliable data center at a lower cost per user. For small or midsize organizations, the ASP can provide greater levels of security and reliability than a customer organization. The ASP obtains this advantage by spreading the costs of innovative solutions over many customers. ASPs also benefit from hiring talented technology experts many small companies cannot afford or would underutilize. The client firm also derives strategic benefits from the ASP relationship in that the ASP keeps it up-to-date on the latest technologies.
The growing popularity of the ASP model has led many organizations to describe themselves as ASPs. Among them are traditional Internet Service Providers (ISPs) marketing themselves as ASPs and adding application-hosting services. Some Web-based companies that offer a single online service are calling themselves ASPs. While running individual applications or services remotely can offer some advantages, much of the user benefit of the ASP model comes from outsourcing computing infrastructure. In the most complete case, an organization is left with just thin clients and on-site networking, with the ASP providing applications and Internet connections.
This article describes the technologies used by ASPs, including infrastructure, security, reliability, and performance issues, as well as the economics of the ASP model from the different stakeholders' perspectives.
ASP infrastructure. In the ASP approach, applications are transmitted to the user machine over a network using a variety of thin client models. A typical model of an ASP is shown in Figure 1. ASPs host applications at the ASP site on servers in a server farm. Applications are dynamically allocated to servers by a switch based on available capacity; servers are not permanently allocated to a particular ASP customer. The application is then transmitted to the customer via the Internet or a private WAN connection, such as Frame Relay or T-carrier.
Two common approaches to the ASP model are the Citrix MetaFrame and the Network Computer (NC)/Java models. The Citrix technology is used for transmitting screen images from a server running an application to a thin client that needs only a small Citrix Independent Computing Architecture (ICA) client. The Citrix Server is most commonly used with MS Windows applications but supports a number of versions of Unix. The data communications between the client and server using the Citrix Server thin client technology are shown in Figure 2. Since the Citrix client runs on many platforms and uses few resources, nearly any machine can act as the thin client, including older DOS-based PCs, Unix workstations, and MS Windows computers. The applications in this architecture are written the same as traditional desktop or client/server applications, so the system supports most legacy systems running on platforms for which the Citrix Server software has been written.
The NC technology uses a variety of Web-related technologies to display and/or execute applications on a thin client [2]. In its most sophisticated configuration, NC systems use the Java model to transmit applications software to the client in bytecode form for client execution. The bytecode is machine-independent code run by the client through a Java virtual machine. Compared to the Citrix architecture, the initial data transfer can be large, as the application is transmitted to the client. However, once loaded in memory, the application does not typically need to communicate with the server. A limitation of this approach is that applications must be written in Java, an uncommon language for legacy systems, but one emerging as a common language for new development. The Java approach, like Citrix MetaFrame, works with most client computer architectures; Java virtual machines have been written for most operating systems and are distributed as open source software. Java virtual machines are now also becoming available for devices, such as phones and PDAs, and is likely to add flexibility for new applications.
An organization relying on an ASP will take the blame from its clients in the event of system failure and therefore needs strong confidence in the ASP.
Both systems allow applications to sit on the desktop alongside traditionally installed applications. The Java applications are run on the client computer and function in the same way as locally housed applications. However, Java applications may have small look and feel differences from local applications, as they are often transported across GUIs. The Citrix client software, in contrast, integrates Citrix applications with the desktop; the user is presented with a seamless user interface when the client and server are running the same operating system. Applications written for the client GUI look the same as local applications, while applications written for other systems retain the look and feel of their native system. A Web browser can run on either system, allowing access to applications using other browser-based user interfaces, even allowing both systems to be used simultaneously.
Both systems can be designed to store data and documents either locally or at the provider, but they differ in terms of which system is local to the application. In the Java architecture, the application runs locally and has faster access to the local hard disk. Data accesses to the provider will be transmitted across the network connection. In the Citrix architecture, the application runs at the provider site, so fast hard disk access will be available to the provider's servers. Access to the local drive is possible but requires transmission across the network.
There are other ways of distributing applications besides the Citrix and Java approaches. They include local area network (LAN) file servers with mapped drives, X Windows, and MS Terminal Services. These systems require higher-bandwidth networks and do not usually work practically over WAN links. Although they provide useful computing models in some situations, they are not being adopted by ASPs.
Security and reliability. Security and reliability are probably the most important performance characteristics of the ASP. Even a small lapse in either will create customer dissatisfaction that is difficult to repair. ASPs generally invest heavily in backup and security systems, including sophisticated personnel controls, backup data systems, computer and network redundancy, and multiple power system contingencies. Personnel controls include fingerprint or IRIS identification, along with passwords and armed guard protection of facilities. Backup data systems include redundant disk drives and off-site tape backup. Computer and network redundancy includes redundant Internet providers, redundant internal networking, and redundant computer systems, as well as full replication of data centers. Multiple power system contingencies include both battery and diesel generator systems. Further improvements are made to power systems by ensuring redundant data centers are on different power grids.
These security and reliability safeguards go beyond what many small to midsize companies can afford and thus are a benefit of the ASP model. On the other hand, some customers will ascribe a significant amount of risk to trusting an external organization with vital data and critical applications. Most organizations are sensitive about their data, but organizations in the medical, legal, and high-tech fields may be especially sensitive. The loss of patient records, legal case records, or high-tech secrets could be devastating to a firm or its clients. An organization relying on an ASP will take the blame from its clients in the event of system failure and therefore needs strong confidence in the ASP.
Such levels of customer trust can be developed through certification by third-party experts. Accounting and consulting firms are offering security audits of ASPs and provide certification to those that pass. This level of certification can improve real as well as perceived levels of security and reliability. However, some certifiers are also ASPs, thus creating potential conflicts of interest. Certifications from ASP suppliers, such as Cisco Systems [1], can be useful, but these supplier certification programs are largely marketing tools that require ASPs to use certain product bundles. Service-level agreements are becoming increasingly sophisticated in helping to define the levels of service promised by the ASP and outlining the types of recourse available to customers if the service level is not met.
Application performance. Both usability and availability are important performance characteristics for ASPs. Whether availability increases or decreases depends on how it is managed. While server farms will typically be more reliable than in-house systems, network outages can render client computers useless. Network contingency plans must therefore be more rigorous than in many organizations that focus on standard PC technology.
Network latency (or delays in the delivery of data) can make the user interface sluggish. For example, quick mouse actions cause the image to lag the user's hand movement. Careful network design can minimize latency and may favor private connections to the provider over ones that rely on the Internet.
The economy-of-scale benefits derived from the model are shared.
ASP Economics
The ASP model affects the economics of all stakeholders, including the user organization and customer, the software developers, and the ASP itself. The ASP can be thought of as playing the role of distributor, much like a distributor of physical goods. This characterization is shown in Figure 3. In physical markets, the distributor allows multiple customers and suppliers to exchange goods and services without the need to have a one-to-one relationship between all customers and providers. In the applications software business, customers usually maintain a relationship with the vendors of many or all of the applications they use. By installing applications on its own server farm, the ASP can effectively outsource part of the internal infrastructure cost, as well as reduce customer-provider communication. Generally this is enhanced when the ASP can provide a range of application needs for the client.
By playing the role of distributor for multiple customers and suppliers, the ASP can realize economies of scale in managing hardware, software, and personnel resources. Hardware is used more efficiently because customers not using the system do not place a load on the server farm. Software licenses can be utilized in a similar way, depending, as discussed later, on the agreement the ASP strikes with the software provider. Because personnel can be more focused, they can be used more effectively and efficiently than in a small organization. The ASP tech support and user support personnel can specialize in an application or a type of technology. By contrast, some small organizations may have a single person responsible for both networks and applications, and others may be unable to cost justify a full-time technical support person.
The economies of scale provided by the ASP as a large-scale distributor accrue to the stakeholders based on the contracts they have negotiated. Often the ASP charges the client a monthly fee based on the number of seats. They can also charge on a metered basis, reflecting actual usage. ASPs pay developers based on the number of licenses, which they may purchase on a one-time or monthly basis, depending on the software provider. Often ASPs are instrumental in changing supplier pricing models from a purchase to a monthly subscription.
The customer organization can potentially reduce IT infrastructure costs by using thin clients (or existing clients) and outsourcing applications to the provider. The customer can pay a monthly amount that provides a savings relative to an in-house solution, while still allowing the ASP to make a profit. Thus, the economy-of-scale benefits derived from the model are shared.
The implication for software companies is less clear, as they may reduce their short-term revenue from software sales. However, if applications are less costly to deploy, it may become more cost effective for organizations to make use of more software. The current competitive pressure forcing down the price of applications is likely to continue. The ASP model facilitates such competition, as ASPs can offer applications vendors large user bases and lower support costs, thus providing an opportunity to negotiate favorable pricing.
Because the ASP usually has more expertise in technology forecasting and a larger customer base, it can usually make better technology investment decisions than customer organizations that are not IT specialists. This brings cost savings, as well as predictability, to customer organizations. The potential benefits of the ASP model are broken out by stakeholder groups in the table here.
Strategic Alliance
The use of an ASP has strategic implications for any organization. Some organizations choose not to use ASPs because they view their information systems as strategic assets. However, ASPs can help organizations maintain a competitive edge by taking over the systems work of the organization and providing the organization with a software development environment for custom applications. In other words, in addition to end-user applications, the ASP can provide the latest application development tools. Furthermore, many organizations will benefit from an alliance with an ASP that has better knowledge of technology trends and that can quickly introduce new technologies to its client organizations.
Another potential strategic benefit of the ASP model is the outsourcing of interorganizational computer systems. Just as the ASP can be a distributor for applications, it can also become a distributor for data, providing an interface between the systems of trading partners. As ASPs use and develop standards for interorganizational computing, they can reduce the cost and time of configuring such systems, leaving their customers to negotiate business deals between partners. By devising new technological standards of business-to-business e-commerce, the ASP changes the process to one where it tackles the technical standards for large numbers of organizations, leaving the organization to tackle the business questions of with whom to share what data.
Conclusion
The ASP concept brings a new configuration of software, computing, and networking from which managers can choose. It promises to increase security, reliability, and availability of a range of software applications at a lower cost per user. In the current tight market for technology professionals, a higher utilization of personnel expertise can benefit the employees, the ASP, and the systems they develop.
This new concept also brings new uncertainty, as it redistributes responsibilities among organizations. Client organizations must have assurances that ASPs can be trusted. Third-party audit firms are developing methods of gauging ASP security while simultaneously becoming ASPs themselves.
Whether or not it brings all of the benefits of older configurations, thereby eclipsing them, remains to be seen; however, it is unlikely it will be the only configuration used by organizations. The PC era ushered in distributed management of computing, which brought with it innovation and rapid response to end-user needs. Some of those benefits are available in the ASP model, but they are likely to remain more restrictive than full-function desktop PCs and departmental file servers.
In the era of computer time-sharing discussed earlier, as with today's ASP era, there are many tough choices to be made. In his 1967 text [5], Zeigler posited several rhetorical questions, including: "How do you know if time-sharing is for you? If it is, what measures do you use for filling your specific requirements?" His advice regarding these issues illustrates there is much we can learn from an earlier generation of computing.
"Evaluate time-sharing computers the way you would any other data processing system. Weigh the advantages against the drawbacks and decide in favor of the balance," Zeigler suggested.
References
1. Cisco Systems. 2001. ASP/AIP business white paper; see www.cisco.com/application/pdf/en/us/guest/partners/pr80/ c1815/ccmigration_09186a00800a2fb8.pdf.
2. Mohseni, P. The Waite Group's NC Guide: Development Strategies for Network Computers. Waite Group Press, Corte Medera, CA, 1997.
3. Network Technology Group, 2000; see www.ntg.com.
4. iManage, 2000. iManage, Inc.; see www.imanage.com.
5. Ziegler, J.R. Time-Sharing Data Processing Systems. Prentice-Hall, Englewood Cliffs, NJ, 1967.
Figures
Figure 1. Typical ASP technology model.
Figure 2. Thin client/application server communication using Citrix Software.
©2003 ACM 0002-0782/03/0800 $5.00
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. To copy otherwise, to republish, to post on servers or to redistribute to lists, requires prior specific permission and/or a fee.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2003 ACM, Inc.
No entries found