1.5. Overview of the Quantitative Approach
Speed, around-the-clock availability, and security are the most common indicators of quality of service of an electronic business site. Management faces a twofold challenge. On one hand, companies must meet customer expectations in terms of quality of service. On the other hand, companies have to keep site costs under control to stay competitive. Therefore, capacity, reliability, scalability, and security are key issues to e-business site managers. E-business sites are complex computer-system architectures, with multiple interconnected layers of software and hardware components, such as networks, caching proxies, routers, high speed links, and mainframes with large databases. The nature of e-business workload is also complex due to its transactional nature, secure requirements, payment protocols, and the unpredictable characteristics of service requests over the Internet. Planning the capacity of e-business sites requires more than just adding extra hardware. It requires more than intuition, ad hoc procedures, and rules of thumb. There are many possible alternative architectures and one has to be able to determine the most cost-effective architecture. This is where the quantitative approach of this book and capacity planning techniques for electronic business come into play.
Figure 1.7 gives an overview of the main steps of the quantitative approach to analyze electronic business sites. The approach is represented by a cycle with a series of steps that are discussed in detail throughout the chapters of this book. The starting point is the business model and its measurable objectives, which are used to establish service level goals and determine what applications are central to the goals. Once the business model and its quantitative objectives have been understood, one is able to go through the quantitative analysis cycle.
Figure 1.7. Quantitative Analysis Cycle of an E-business Site.
1. |
The first step entails obtaining an in-depth understanding of the e-business site architecture. What are the system requirements of the business model? What is the configuration of the site in terms of servers and internal connectivity? How many internal layers are there in the site? What types of servers (i.e., HTTP, database, or authentication) is the site running? What type of software (i.e., operating system, HTTP server software, transaction monitor, DBMS) is used in each server machine? How reliable and scalable is the architecture? This step should yield a systematic description of the website environment, its components, and services. |
2. |
The second step consists of measuring the performance of the systems that make up the e-business site. This is a key step in the process of guaranteeing quality of service and preventing problems. Performance measurements should be collected from different reference points, each point carefully chosen to observe and monitor the environment under study. For example, logs of transactions and accesses to servers are the main source of information. Further information, such as downloading times from different points of the network, may help to follow the service level perceived by customers. The information collected should be capable of answering questions like the following. What is the number of customer visits per day? What is the site revenue for a specific period of time? What is the average and peak traffic to the site? What characterizes the shoppers of a particular set of products? |
3. |
The third step focuses on understanding the customer behavior. Customers interact with an electronic business site through sequences of requests to invoke the various services available in the site. Some customers may be considered as heavy buyers while others, considered occasional buyers, spend most of their time browsing and searching the site. These two examples of classes of customers exhibit different navigational patterns and, as a consequence, invoke services in different ways with different frequencies. The point is that each service may exercise the site's resources in a different manner. Some services may use a large amount of processing time from the application server while others may concentrate on the database server. Understanding customer behavior is critical for achieving the business objectives as well as for adequately sizing the resources of the site. |
4. | The fourth step characterizes the workload of an e-business site. E-business workloads are composed of sessions, sequences of requests of different types made by a single customer during a single visit to a site. Examples of requests from an online shopper are browse, search, select, add to the shopping cart, user registration, and pay. An online trader would have different operations, such as enter a stock order, research a fund, obtain real-time quotes, retrieve company profiles, and compute earning estimates. This step uses the representations generated by the previous step. Another aspect of the nature of the e-business workload is its intensity, which measures how often customers arrive to the site. |
5. | In the fifth step, quantitative techniques and analytical models based on queuing network theory are used to evaluate the performance of electronic business sites. Performance models can be used to predict performance when any aspect of the workload or the site architecture is changed. |
6. | The sixth step aims at obtaining the input parameters for performance models. These parameters describe the architecture of the e-business site and the workload under study. A site architecture typically includes Web servers, application servers, and database servers organized in layered groups. Each server consists of many basic components that contribute to the system performance, such as processors, disks, and network interfaces. Some parameters can be directly derived from measurement data while others must be estimated according to specific features of the site. |
7. | The seventh step forecasts the expected workload for an e-business site. The techniques and strategies for forecasting demand in the electronic marketplace should provide answers to questions: What is the expected workload of the e-business website of an electronic bookstore during the holiday season? How will the number of users of an online auction vary during the next six months? |
8. | Using the performance models and workload forecasts, the last step of the cycle analyzes many possible alternative architectures to determine the most cost-effective one. The future scenarios should take into consideration the expected workload, the site cost, and the quality of service perceived by customers. |