Client/Server Software Testing
By: Hongmei Yang
I: Introduction to Client/Server architecture:
Client/Server system development is the preferred method of constructing cost-effective department- and enterprise-level strategic corporate information systems. It allows the rapid deployment of information systems in end-user environments.
1: What is Client/Server Computing?
Client/Server computing is a style of computing involving multiple processors, one of which is typically a workstation and across which a single business transaction is completed .
Client/Server computing recognizes that business users, and not a mainframe, are the center of a business. Therefore, Client/Server is also called “client-centric” computing.
Today, Client/Server computing is extended to the Internet—netcentric computing (network centric computing), the concepts of business users have expanded greatly. Forrester Report describes the netcentric computing as “Remote servers and clients cooperating over the Internet to do work” and says that Internet Computing extends and improves the Client/Server model .
The characteristics of Client/Server computing includes:
1. There are multiple processors.
2. A complete business transaction is processed across multiple servers
Netcentric computing ---- as an evolution of Client/Server model, has brought new technology to the forefront, especially in the area of external presence and access, ease of distribution, and media capabilities. Some of new technologies are :
Browser, which provides a “universal client”. In the traditional Client/Server environment, distributing an application internally or externally for an enterprise requires that the application be recompiled and tested for all specific workstation platforms (operating systems). It also usually requires loading the application on each client machine. The browser-centric application style offers an alternative to this traditional problem. The web browser provides a universal client that offers users a consistent and familiar user interface. Using a browser, a user can launch many types of applications and view many types of documents. This can be accomplished on different operating systems and is independent of where the applications or documents reside.
Direct supplier-to-customer relationships. The external presence and access enabled by connecting a business node to the Internet has opened up a series of opportunities to reach an audience outside a company’s traditional internal users.
Richer documents. Netcentric technologies (such as HTML, documents, plug-ins, and Java) and standardization of media information formats enable support for complex documents, applications and even nondiscrete data types such as audio and video.
Application version checking and dynamic update. The configuration management of traditional Client/Server applications, which tend to be stored on both the client and server sides, is a major issue for many corporations. Netcentric computing can checking and update application versions dynamically.
2: Architectures for Client/Server System.
Both traditional Client/Server as well as netcentric computing are tiered architectures. In both cases, there is a distribution of presentation services, application code, and data across clients and servers. In both cases, there is a networking protocol that is used for communication between clients and servers. In both cases, they support a style of computing where processes on different machines communicate using messages. In this style, the “client” delegates business functions or other tasks (such as data manipulation logic) to one or more server processes. Server processes respond to messages from clients.
A Client/Server system has several layers, which can be visualized in either a conceptual or a physical manner. Viewed conceptually, the layers are presentation, process, and database. Viewed physically, the layers are server, client, middleware, and network.
2.1. Client/Server 2-tiered architecture:
2-tiered architecture is also known as the client-centric model, which implements a “fat” client. Nearly all of the processing happens on the client, and client accesses the database directly rather than through any middleware. In this model, all of the presentation logic and the business logic are implemented as processes on the client.
2-tiered architecture is the simplest one to implement. Hence, it is the simplest one to test. Also, it is the most stable form of Client/Server implementation, making most of the errors that testers find independent of the implementation. Direct access to the database makes it simpler to verify the test results.
The disadvantage of this model is the limit of the scalability and difficulties for maintenance. Because it doesn’t partition the application logic very well, changes require reinstallation of the software on all of the client desktops.
2.2. Modified 2-tiered architecture:
Because of the nightmare of maintenance of the 2-tiered Client/Server architecture, the business logic is moved to the database side, implemented using triggers and procedures. This kind of model is known as modified 2-tiered architecture.
In terms of software testing, modified 2-tiered architecture is more complex than 2-tiered architecture for the following reasons:
It is difficult to create a direct test of the business logic. Special tools are required to implement and verify the tests.
It is possible to test the business logic from the GUI, but there is no way to determine the numbers of procedures and/or triggers that fires and create intermediate results before the end product is achieved.
Another complication is dynamic database queries. They are constructed by the application and exist only when the program needs them. It is very difficult to be sure that the test generates a query “correctly”, or as expected. Special utilities that show what is running in memory must be used during the tests.
2.3. 3-tiered architecture:
For 3-tiered architecture, the application is divided into a presentation tier, a middle tier, and a data tier. The middle tier is composed of one or more application servers distributed across one or more physical machines. This architecture is also termed the “the thin client—fat server” approach.
This model is very complicated for testing because the business and/or data objects can be invoked from many clients, and the objects can be partitioned across many servers. The characteristics make the 3-tiered architecture desirable as a development and implementation framework at the same time make testing more complicated and tricky.
3: Critical Issues Involved in Client/Server System Management:
Hurwitz Consulting Group, Inc. has provided a framework for managing Client/Server systems that identifies eight primary management issues :
Configuration and administration
Data and storage
II Client/Server Software Testing:
Software testing for Client/Server systems (Desktop or Webtop) presents a new set of testing problems, but it also includes the more traditional problems testers have always faced in the mainframe world. Atre describes the special requirements of Client/Server testing :
The client’s user interface
The client’s interface with the server
The server’s functionality
The network (the reliability and performance of the network)
Introduction to the Client/Server Software Testing:
We can view the Client/Server software testing from different perspectives:
From a “distributed processing” perspective: Since Client/Server is a form of distributed processing, it is necessary to consider its testing implication from that point of view. The term “distributed” implies that data and processes are dispersed across various and miscellaneous platforms. Binder states several issues needed to be considered in the Client/Server environments .
Client GUI considerations
Target environment and platform diversity considerations
Distributed database considerations (including replicated data)
Distributed processing considerations (including replicated processes)
Nonrobust target environment
Nonlinear performance relationships
From a cross-platform perspective: The networked cross-platform nature of Client/Server systems requires that we pay much more attention to configuration testing and compatibility testing. The purpose of configuration testing is to uncover the weakness of the system operated in the different known hardware and software environments. The purpose of comparability testing is to find any functionally inconsistency of the interface across hardware and software.
From a cross-window perspective: The current proliferation of Microsoft Windows environments has created a number of problems for Client/Server developers. For example, Windows 3.1 is a 16-bit environment, and Window 95 and Window NT are 32-bit environment. Mixing and matching 16- bit and 32-bit code/16bits or 32bits systems and products causes major problems. Now there exit some automated tools that can generate both 16-bit and 32-bit test scripts.
2. Testing Plan for Client/Server Computing:
In many instances, testing Client/Server software cannot be planned from the perspective of traditional integrated testing activities because this view either is not applicable at all or is too narrow, and other dimensions must be considered. The following are some specific considerations needing to be addressed in a Client/Server testing plan.
Must include consideration of the different hardware and software platforms on which the system will be used.
Must take into account network and database server performance issues with which mainframe systems did not have to deal.
Has to consider the replication of data and processes across networked servers
See attached “Client/Server test plan based on application functionality” .
In the test plan, we may address or construct several different kinds of testing:
The system test plan: System test scenarios are a set of test scripts, which reflect user behaviors in a typical business situation. It’s very important to identify the business scenarios before constructing the system test plan.
See attached CASE STUDY: The business scenarios for the MFS imaging system
The user acceptance test plan: The user acceptance test plan is very similar to the system test plan. The major difference is direction. The user acceptance test is designed to demonstrate the major system features to the user as opposed to finding new errors.
See attached CASE STUDY: Acceptance test specification for the MFS imaging system
The operational test plan: It guides the single user testing of the graphical user interface and of the system function. This plan should be constructed according to subsection A and B of Section II in the testing plan template -- Client/Server test plan based on application functionality. (See attached Appendix I)
The regression test plan: The regression test plan occurs at two levels. In Client/Server development, regression testing happens between builds. Between system releases, regression testing also occurs postproduction. Each new build/release must be tested for three aspects:
To uncover errors introduced by the fix into previously correct function.
To uncover previously reported errors that remain.
To uncover errors in the new functionality.
Multiuser performance test plan: It is necessary to be performed in order to uncover any unexpected system performance problem under load. This test plan should be constructed form Section V of the testing plan template-- Client/Server test plan based on application functionality. (See attached Appendix I)
3. Client/Server Testing in Different Layers:
3.1. Testing on the Client Side—Graphic User Interface Testing:
3.1.1 The complexity for Graphic User Interface Testing is due to:
Cross-platform nature: The same GUI objects may be required to run transparently (provide a consistent interface across platforms, with the cross-platform nature unknown to the user) on different hardware and software platforms
Event-driven nature: GUI-base applications have increased testing requirements because they are in an event-driven environment where user actions are events that determine the application’s behavior. Because the number of available user actions is very high, the number of logical paths in the supporting program code is also very high.
The mouse, as an alternate method of input, also raises some problems. It is necessary to assure that the application handles both mouse input and keyboard input correctly.
The GUI testing also requires testing for the existence of a file that provides supporting data/information for text objects. The application must be sensitive to the existence, or nonexistence.
In many cases, GUI testing also involves the testing of the function that allows end-users to customize GUI objects. Many GUI development tools give the users the ability to define their own GUI objects. The ability to do this requires the underlying application to be able to recognize and process events related to these custom objects.
3.1.2 GUI testing techniques: Many traditional software testing techniques can be used in GUI testing.
Review techniques such as walkthroughs and inspections . These human testing procedures have been found to be very effective in the prevention and early correction of errors. It has been documented that two-thirds of all of the errors in finished information systems are the results of logic flaws rather than poor coding . Preventive testing approaches, such as walkthroughs and inspections can eliminate the majority of these analysis and design errors before they go through to the production system.
Data validation techniques: Some of the most serious errors in software systems have been the result of inadequate or missing input validation procedures. Software testing has powerful data validation procedures in the form of the Black Box techniques of Equivalence Partitioning, Boundary Analysis, and Error Guessing. These techniques are also very useful in GUI testing.
Scenario testing: It is a system-level Black Box approach that also assure good White Box logic-level coverage for Client/Server systems.
The decision logic table (DLT): DLT represents an external view of the functional specification that can be used to supplement scenario testing from a logic-coverage perspective. In DLTs, each logical condition in the specification becomes a control path in the finished system. Each rule in the table describes a specific instance of a pathway that must be implemented. Hence, test cases based on the rules in a DLT provide adequate coverage of the module’s logic independent of its coded implementation.
In addition to these traditional testing techniques, a number of companies have begun producing structured capture/playback testing tools that address the unique properties of GUIs. The difference between traditional capture/playback and structured capture/playback paradigm is that capture/playback occurs at an external level. It records input as keystrokes or mouse actions and output as screen images that are saved and compared against inputs and output images of subsequent tasks.
Structured capture/playback is based on an internal view of external activities. The application program’s interactions with the GUI are recorded as internal ‘events” that can be saved as “scripts” written in some certain language.
3.2 Testing on the Server Side---Application Testing:
There are several situations that scripts can be designed to invoke during several tests: load testing, volume tests, stress tests, performance tests, and data-recovery tests.
3.2.1 Client/Server loading tests:
Client/Server systems must undergo two types of testing: single-user-functional-based testing and multiuser loading testing.
Multiuser loading testing is the best method to gauge Client/Server performance. It is necessary in order to determine the suitability of application server, database server, and web server performance. Because multiuser load test requires emulating a situation in which multiple clients access a single server application, it is almost impossible to be done without automation.
For the Client/Server load testing, some common objectives include:
Measuring the length of time to complete an entire task
Discovering which hardware/software configuration provides optimal performance
Tuning database queries for optimal response
Capturing Mean-Time-To-Failure as a measure of reliability
Measuring system capacity to handle loads without performance degradation
Identifying performance bottlenecks
Based on the test objectives, a set of performance measurements should be described. Typical measurements include:
End-to-end response time
Network response time
GUI response time
Server response time
Middleware response time
3.2.2 Volume testing:
The purpose of volume testing is to find weaknesses in the system with respect to its handling of large amount of data during extended time periods
3.2.3 Stress testing:
The purpose of stress testing is to find defects of the system capacity of handling large numbers of transactions during peak periods. For example, a script might require users to login and proceed with their daily activities while, at the same time, requiring that a series of workstations emulating a large number of other systems are running recorded scripts that add, update, or delete from the database.
3.2.4 Performance testing:
System performance is generally assessed in terms of response time and throughput rates under differing processing and configuration conditions. To attack the performance problems, there are several questions should be asked first:
How much application logic should be remotely executed?
How much updating should be done to the database server over the network from the client workstation?
How much data should be sent to each in each transaction?
According to Hamilton , the performance problems are most often the result of the client or server being configured inappropriately.
The best strategy for improving client-sever performance is a three-step process . First, execute controlled performance tests that collect the data about volume, stress, and loading tests. Second, analyze the collected data. Third, examine and tune the database queries and, if necessary, provide temporary data storage on the client while the application is executing.
3.2.5 Other server side testing related to data storage:
Data recovery testing
Data backup and restoring testing
Data security testing
Replicated data integrity testing.
3.2.6 Examples for automated server testing tools:
LoadRunning/XL, offered from Mercury Interactive, is a Unix-based automated server testing tool that tests the server side of multiuser Client/Server application. LoadRunning/PC is similar to products based on Windows environments.
SQL Inspector and ODBC Inspector are tools for testing the link between the client and the server. These products monitor the database interface pipeline and collect information about all database calls or a selected subset of them.
SQL Profiler, is used for tuning database calls. It stores and displays statistics about SQL commands embedded in Client/Server applications.
SQLEYE is an NT-based tool, offered by Microsoft. It can track the information passed through the SQL Server and its client. Client application connect indirectly to SQL server through SQLEYE, which allows users to view the queries sent to SQL Server, the returned results, row counts, message, and errors
3.3 Networked Application Testing
Testing the network is beyond the scope of an individual Client/Server project as it may serve more than a single Client/Server project. Thus, network testing falls into the domain of the network management group. As Robert Buchanan  said: “If you haven’t tested a network solution, it’s hard to say if it works. It may ‘work’. It may execute all commands, but it may be too slow for your needs”.
Nemzom blames the majority of network performance problem on insufficient network capacity . He views bandwidth and latency as the critical determinants of network speed and capacity. He also sees interactions among intermediate network nodes (switches, bridges, routers and gateways) as adding to the problem.
Elements of network testing include:
Application response time measures
Throughput and performance measurement
Configuration and sizing
Stress testing and performance testing
It is necessary to measure application response time while the application is completing a series of tasks. This kind of measure reflects the user’s perception of the network, and is applicable through the entire network life cycle phase. Testing application functionality involves testing shared functionality across workstations, shared data, and shared processes. This type of testing is applicable during the development and evolution. Configuration and sizing measure the response of specific system configurations. This is done for different network configurations until the desired performance level is reached. . The point of stress testing is to overload network resource such as routers or hubs. Performance testing can be used to determine how many network devices will be required to meet the network’s performance requirements. Reliability testing involves running the network for 24-72 hrs under a medium-to-heavy load. From a reliability point of view, it is important that the network remain functional in the event of a node failure.
4 Special Concerns for Internet Computing --- Security Testing:
For internet-based Client/Server systems, security testing for the web server is important. The web server is your LAN’s window to the world and, conversely, is the world’s window to your LAN.
The following excerpt is taken from the WWW Security FAQ :
It’s a maxim in system security circles that buggy software opens up security holes. It’s a maxim in software development circles that large, complex programs contain bugs. Unfortunately, web servers are large, complex programs that can contain security holes. Furthermore, the open architecture of web server allows arbitrary CGI scripts to be executed on the server’s side of the connection in response to remote requests. Any CGI script installed at your site may contain bugs, and every such bug is a potential security hole.
Three types of security risks have been identified :
The primary risk is errors in the web server side misconfiguration that would allow remote users to:
Steal confidential information
Execute commands on the server host, thus allowing the users to modify the system
Gain information about the server host that would allow them to break into the system
Launch attacks that will bring the system down.
The secondary risk occurs on the Browser-side
Active content that crashes the browser, damages your system, breaches your company’s privacy, or creates an annoyance.
The misuse of personal information provided by the end user.
The tertiary risk is data interception during data transfer.
The above risks are also the focuses of web server security testing. As a tester, it is your responsibility to test if the security extends provided by the server meet the user’s expectation for the network security.
Client/Server system development is the preferred method of constructing cost-effective department- and enterprise-level strategic corporate information systems. It allows the rapid deployment of information systems in end-user environments
Both traditional Client/Server as well as netcentric computing are tiered architectures. Currently, the dominant three types of Client/Server architectures include 2-tiered architecture, modified 2-tiered architecture, and three tiered architecture. 2-tiered architecture is the simplest one to implement, and the simplest one to test. The characteristics of the 3-tiered architecture that make desirable as development and implementation framework at the same time make testing more complicated
Testing Client/Server software cannot be planned from the perspective of traditional integrated testing activities. In a Client/Server testing plan, some specific considerations, such as different hardware and software platforms, network and database server performance issues, the replication of data and processes across networked servers, etc. need to be addressed.
The complexity for GUI (Graphic User Interface) testing is increase because of some characteristics of GUIs, for instance, its cross-platform nature, event-driven nature, and an additional input method—mouse. Many traditional software testing techniques can be used in GUI testing. Currently, a number of companies have begun producing structured capture/playback tools that address the unique properties of GUIs.
There are several situations that scripts can be designed to be invoked during server tests: load testing, volume tests, stress tests, performance tests, and data-recovery tests. These types of testing are nearly impossible without automation. Some sophisticated testing tools used in server side testing already emerged in the market, such as LoadRunning/Xl, SQL Inspector, SQL profiler, and SQLEYE.
Network test is a necessary but difficult series of tasks. Its difficulty is compounded by the fact that Client/Server development may be targeted for an exiting network or for one that is yet to be installed. Proactive network management and proper capacity planning will be very helpful. In addition, performance and stress testing can ease the network testing burden.
For internet-based Client/Server systems, security testing for the web server is important. The web server is your LAN’s window to the world and, conversely, is the world’s window to your LAN. As a tester, it is your responsibility to find weakness in the system security
... to read more articles, visit http://sqa.fyicenter.com/art/