Trends in Software Testing

As the complexity of software applications increases, testing becomes more crucial. And in the process, more time consuming. Here is a list of emerging testing practices.

Software is everywhere today and is becoming increasingly mission critical, whether in satellites and planes, or e-commerce websites. Software complexity is also on the rise - thanks to distributed, multi-tier applications targeting multiple devices (mobile, thin/thick clients, clouds, etc). Added to that are development methodologies like extreme programming and agile development. No wonder software testing professionals are finding it hard to keep up with the change.

As a result, many projects fail while the rest are completed significantly late, and provide only a subset of the originally planned functionality. Poorly tested software and buggy code cost corporations billions of dollars annually, and most defects are found by end users in production environments.

Given the magnitude of the problem, software-testing professionals are finding innovative means of keeping up - both in terms of tools and methodologies. This article covers some of the recent trends in software testing - and why they're making the headlines. Test driven development (TDD)

TDD is a software development technique that ensures your source code is thoroughly unit-tested as compared to traditional testing methodologies, where unit testing is recommended but not enforced. It combines test-first development (where you write a test before you write just enough code to fulfil that test), and refactoring (where, if the existing design isn't the best possible to enable you to implement a particular functionality, you improve it to enable the new feature).

TDD is not a new technique-but it is suddenly centre stage, thanks to the continued popularity of software development methodologies such as agile development and extreme programming.

Optimisations to TDD include the use of tools (such as PEX/peer exchange for Visual studio - http://research.microsoft.com/en-us/projects/pex/ ) to improve code coverage, by creating parameterised unit tests that look for boundary conditions, exceptions, and assertion failures.

TDD is gaining popularity as it allows for incremental software development - where bugs are detected and fixed as soon as the code is written, rather than at the end of an iteration or a milestone.

For more details on TDD, use the following links:
http://en.wikipedia.org/wiki/Test-driven_development
http://www.agiledata.org/essays/tdd.html

Virtualisation testing:-

Testing is becoming increasingly complex - the test environment set-up, getting people access to the environment, and loading it with the right bits from development, all take up about 30-50 per cent of the total testing time in a typical organisation. What is worse is that when testers find bugs, it is hard to re-create the same environment for developers to investigate and fix bugs. Test organisations are increasingly gravitating towards virtualisation technologies to cut down test set-up times significantly. These technologies include:

• accelerate set-up/tear down and restoration of complex virtual environments to a clean state, improving machine utilisation

• eliminate no repro bugs by allowing developers to recreate complex environments easily

• improve quality by automating virtual machine provisioning, building deployment, and building verification testing in an integrated manner (details later)

As an offshoot, virtualisation ensures that test labs reduce their energy footprint, resulting in a positive environmental impact, as well as significant savings.

Some of the companies that have virtual test lab management solutions are VMware, VMLogix, and Surgient. Microsoft has recently announced a Lab Management (http://channel9.msdn.com/posts/VisualStudio/Lab-Management-coming-to-Visual-Studio-Team-System-2010/) product as part of its Visual Studio Team System 2010 release. Lab Management supports multiple environment management, snapshots to easily restore to a previous state, virtual network isolation to allow multiple test environments to run concurrently, and a workflow to allow developers to have easy access to environments to reproduce and fix defects.

Theresa Lanowitz, founder of Voke, a firm involved with analysis of trends in the IT world, expects virtualisation to become ‘the defining technology of the 21st century', with organisations of every size set to benefit from virtualisation as a part of its core infrastructure.

Continuous integration:-

CI is a trend that is rapidly being adopted in testing, where the team members integrate their work with the rest of the development team on a frequent basis by committing all changes to a central versioning system. Beyond maintaining a common code repository, other characteristics of a CI environment include build automation, auto-deployment of the build into a production-like environment, and ensuring a self-test mechanism such that at the very least, a minimal set of tests are run to confirm that the code behaves as expected.

Leveraging virtualised test environments, tools such as Microsoft's Visual Studio Team System (VSTS) can create sophisticated CI workflows. As soon as code is checked in, a build workflow kicks in that compiles the code - deploys it on to a virtualised test environment, triggers a set of unit and functional tests on the test environment, and reports on the results.

VSTS takes the build workflow one step further, and performs the build before the check-in is finalised, allowing the check-in to be aborted if it would cause a break, or if it fails the tests. And given historical code coverage data from test runs, the tool can identify which one of the several thousand test cases needs to be run when a new build comes out - significantly reducing the build validation time.

One obvious benefit of continuous integration is transparency. Failed builds and tests are found quickly rather than having to wait for the next build. The developer who checked in the offending code is probably still nearby and can quickly fix or roll back the change.

For a complete set of tools that help enable CI, see http://en.wikipedia.org/wiki/Continuous_Integration

Crowd testing:-

Crowd testing is a new and emerging trend in which, rather than relying on a dedicated team of testers (in-house or out sourced), companies rely on virtual test teams (created on demand) to get complete test coverage and reduce the time to market for their applications.

The company defines its test requirements in terms of scenarios, environments, and the type of testing (functional, performance, etc). A crowd test vendor (such as uTest - www.utest.com) identifies a pool of testers that meet the requirements, creates a project, and assigns work. Testers check the application, report bugs, and communicate with the company via an online portal. Crowd testing vendors also provide other tools, such as powerful reporting engines and test optimisation utilities. Some of the crowd testing vendors are domain specific - such as Mob4hire (www.mob4hire.com), which focuses on mobile application testing. Testers will bid on various projects specific to their handsets. Developers will choose the testers that they require, and will deploy test plans for the mobile application they are developing. On completion of the test, the mobile tester will get paid for the work.

One obvious advantage is in terms of reducing the test cycle time. But crowd testing is being used in various other scenarios as well - for example, to do usability studies on new user interfaces. The cost savings can be substantial.

Tools driven developer testing:-

Traditionally, developer testing was primarily limited to unit testing and some code coverage metrics. However, as organisations realised that the cost of defects found in development was exponentially lower than that found in test or production, they have begun to invest in tooling to enable developers to find bugs early on.

IDE-integrated tools have made the self-testing practice acceptable to developers, and the unit-testing and coverage analysis process automated for them. These tools also make it easy to analyse performance and compare it with a baseline by extending the unit test infrastructure.

Development teams are also expected to perform a level of security testing (threat modelling, buffer overflow, sequel injection, etc). For teams developing on native languages such as C/C++, developers are also required to use run-time analysis tools to check for memory leaks, memory corruptions and thread deadlocks. Developers are also using static analysis tools to find accessibility, localisation and globalisation issues -- and in some cases more sophisticated errors related to memory management and performance simulation -- by using data flow analysis and other techniques.

As a result of using these innovative methods, testers can now spend a lot more of their time on integration testing, stress, platform coverage, and end-to-end scenario testing. This will help them detect higher-level defects that would have otherwise trickled down to production.

Source: IT Magazine

Twill - Web Automation Tool in Python

Twill is a simple language that allows users to browse the Web from a command-line interface. With twill, you can navigate through Web sites that use forms, cookies, and most standard Web features. Twill supports automated Web testing and has a simple Python interface. Twill is open source and written in Python. Unlike many test automation tools for web application testing, Twill does not launch or need browser. Twill works from the command line (or from your Python script) and allows you to perform standard operations like navigating to specific pages, use forms, cookies and so on. As mentioned earlier, Twill does not launch real browser and hence its usage as test automation tool is limited. Twill can not be used to test rendering of pages, functionality related to Java scripts and so on. Tools like Selenium are better suited for that purpose. Twill is better suited for validating functionality which can be exercised using http request / response and analyzing page source.

Download link: http://darcs.idyll.org/~t/projects/twill-0.9.tar.gz

You can also use Python's easy_install to install or upgrade twill. It works with Python 2.3 or later.

To start using twill, install it and then type twill-sh. At the prompt type:

go http://www.slashdot.org/
show
showforms
showhistory

Source: Testing Site

Selenium IDE - Introduction

Selenium is an open source tool for web application testing. This tool is primarily developed in Java Script and browser technologies and hence supports all the major browsers on all the platforms. For example, you can have your automation scripts written for Firefox on Windows and run them on Firefox in Mac. Most of the time, you will not need to change your scripts for them to work on Mac. In terms of coverage for platform and browser, Selenium is probably one of the best tool available in the market for web applications. There are three variants of Selenium, which can be used in isolation or in combination to create complete automation suite for your web applications. Selenium IDE Selenium Core Selenium Remote Control In this article, we will discuss Selenium IDE. Subsequent articles in the series will cover Selenium Remote Control and Selenium Core as well. Selenium IDE Selenium IDE is the easiest way to use Selenium and most of the time it also serves as a starting point for your automation. Selenium IDE comes as an extension to the Firefox web browser. This can be installed from either openqa or mozilla distribution site. Selenium extension will be downloaded as XPI file. If you open this file using File -> open in Mozilla, it should get installed. Biggest drawback of Selenium IDE is its limitation in terms of browser support. Though Selenium scripts can be used for most of the browser and operating system, Scripts written using Selenium IDE can be used for only Firefox browser if it is not used with Selenium RC or Selenium Core. Selenium IDE is the only flavor of Selenium which allows you to record user action on browser window. It can also record user actions in most of the popular languages like Java, C#, Perl, Ruby etc. This eliminates the need of learning new vendor scripting language. For executing scripts created in these languages, you will need to use Selenium Remote Control. If you do not want to use Remote Control than you will need to create your test scripts in HTML format. So if you are excited about the tool, Lets start playing with Selenium IDE now. If installed properly, Selenium can be accessed from tool --> Selenium IDE in your browser toolbar. As compared to most of the test automation tools it is very simple and lightweight. The small red button on the right hand side gives you an indication on whether Selenium is in recording mode or not. Also, Selenium IDE will not record any operation that you do on your computer apart from the events on Firefox browser window. So go ahead read your mail, open a word doc or do anything else, Selenium will record only your actions on browser. If you are curious to know about other options present on the Selenium IDE, there are not much :) . Other options present on the Selenium IDE toolbar are related to test execution. Run will execute the tests with the maximum possible speed, Walk will execute them with relatively slow speed and in step mode you will need to tell Selenium to take small steps. Final button present on the Selenium IDE toolbar is the Selenium TestRunner. Test Runner gives you nice browser interface to execute your tests and also gives summary of how many tests were executed, how many passed and failed. It also gives similar information on commands which were passed or failed. TestRunner is also available to tests developed in HTML Only. If you open the option window by going to Option , you will see there are some self explanatory options available. For example, encoding of test files, timeout etc. You can also specify Selenium Core and Selenium IDE extensions on this page. Selenium extensions can be used to enhance the functionality provided by Selenium. Selenium extensions are not covered in this article, there will be a separate article for specifying and developing extensions for Selenium...

Test Automation [Key Factors]

Based on the test automation experience the followings are the key factors contributed to a successful test automation project:

1] Dedication to automation (Instead of treating it as a spare-time activity)

2] Commitment by the entire team (rather than just one or two testers)

3] Commitment to automation from the start (rather than trying to automate a manual process later)

4] Making use of correct tools/frameworks/technology

5] Allocating sufficient time and resources

Source: SoftwareQA Blogspot

QA Outsourcing [Reduce Testing Cost]

There are all manner of outsourcing firms who can provide you with impressive powerpoint presentations showing how you can cut your QA costs by 30% or more by outsourcing the work. I would caution you to consider the source however.

Of course Outsourcing firms will produce evidence showing why their service is a valuable cost savings, they will even have little problem letting the managers and executives who decided to use them generate the metrics that show how much of a cost savings it was to outsource. In the case of the Outsourcing firm, it is little more than marketing speak trying to generate new business, expecting them to say any different would be like expecting a Cigarette company executive to go on Television and say "Of course Cigarettes kill people but as long as they are willing to buy them we'll be happy to take their money". As far as their Customers, well admitting that a decision to outsource was a failure would effectively end most executive's or manager's careers so they have little incentive to look too critically at the numbers.

So how much is the cost savings?

That depends on a LOT of factors. First, how much do your inhouse testers cost? Not just their salaries but their total cost.

In the US and UK the total cost of an employee is considered to be around double their salary (that varies some based on the level of benefits offered, location, and other similar factors) and QA testers earn between $45,000 and $90,000 a year depending on experience level, industry, and location. This would mean that on average a Tester in the US should have an employment cost somewhere around $60 - $75 per hour. Only you would have any clue what your costs would be, I'm just including that as a reference point.

The Bill Rates for Outsourced testers ranges from $25 - $40 per hour in places like India and China up to as much as $60 an hour for outsourced testers in industrialized countries (and it can be MUCH higher than that in some locations of if you require specialized knowledge, I've personally seen them as high as $150 per hour).

On top of that bill rate from the outsourcing firm there are additional costs that would work out to an extra $5 - $10 per hour to cover the legal expenses of negotiating the contract with them, the necessary network infrastructure work to allow the outsourcing vendor's network interface with yours securely and so on.

So, a reasonable projection is that in most cases Outsourcing IT work will cost around Half of what it cost to do the work in house on a per hour basis (at least for companies based in the US).

We're not at the end of the caluclation yet though. See that is a Per Hour cost. It assumes that the Outsourced testers are able to work exactly as efficiently as your inhouse testers and with the same skill. While the question of tester skills is highly random from company to company the question of efficiency is actually fairly static. Simply moving your testers into a different building within the same complex creates enough inefficiency to make testing projects take 25 to 50% longer than they would have if they sat next to the developers. Adding in different corporate cultures or even worse national cultures, language barriers, time zone differientals, etc. and the man hour increase is closer to an additional 100% to as much as 200%.

So a 50% reduction in cost per man hour but it takes you 2 to 3 times as many man hours says that Outsourcing is anything from a break even to a money loosing proposition until you factor in variability in skill levels between your existing test staff and the staff at whoever you outsource to. If you have very weak testers and are unable to recruit better ones the it may be that outsourcing will produce a small cost savings of maybe 10 to 20% of your testing costs.

I have yet to encounter an Outsourced testing situation where it actually saved money. Many VP's, Director's, and Manager's claimed it did on the basis of reduced per hour labor costs but never one that was a net positive to the company once all of the outsourcing costs were factored in.

Source: SoftwareQA Blogspot