Jumpstarting DevOps With Continuous Testing

Cognizant 20-20 InsightsJumpstarting DevOps withContinuous TestingBy tightly connecting testing with development and operations,and automating the design, development, quality assuranceand deployment of new applications and systems, IT organizationscan more effectively collaborate and deliver on the dualmandate of increasing operational agility and speeding software’stime to market.Executive SummaryIT organizations are adopting DevOps practicesto enable more frequent software releasesto production, both to accelerate time tomarket and improve the overall customerexperience. But adopting DevOps practicesrequires a cultural shift in terms of the people,processes and technologies involved in development, testing and operations. Many organizations encounter significant challenges whenembracing DevOps, particularly when implementing it in legacy areas. Such obstacleshave proved far more difficult to overcomecompared with those encountered by digitalnative companies.A popular belief is that DevOps requires throwingaway everything the organization has built andstarting from scratch. However, that is untrue;cognizant 20-20 insights december 2015the critical component of DevOps in large organizations is the need to improve communicationand collaboration between IT team disciplines.Enabling such behavior can be cumbersome,especially when regulatory and audit compliancerequires existing organizational structures toremain as-is.In traditional IT organizations, developers,testers and operations engineers typically playdifferent roles, with varying responsibilities,job descriptions and cultures, and all workingin separate functional areas as distinct entities.The question is whether QA can unify thesevarious groups under the DevOps flag andspur adoption. This white paper will attemptto answer that question and elaborate on theapproach.

The Role of Testing in DevOpsThe word DevOps was coined by combining thedevelopment and operations functions. DevOpsis not a technology, a process or a standard;rather, it is an IT culture or movement thatemphasizes ways in which development, testingand operations can collaborate more effectively (see Figure 1). DevOps is more about trust,people and teamwork than about process, andcreating software as an ongoing service, not astatic product.1While there are several interpretations of theDevOps definition, all share the same workingunderstanding. Gartner, for instance, definesDevOps as ” a change in IT culture, focusing onrapid IT service delivery through the adoptionDevOps: An Intersectionof Functionsof agile, lean practices in the context of asystem-oriented approach. DevOps emphasizespeople (and culture), and seeks to improve collaboration between operations and development teams. DevOps implementations utilizetechnology — especially automation toolsthat can leverage an increasingly programmable and dynamic infrastructure from a lifecycleperspective.”2DevOps has emerged for the following reasons: The existence of traditionally siloed teamstructures that did not scale to meet the variedneeds of modern enterprises. The philosophical disconnect between development, testing and operations teams, whichresulted in poor communication, collaborationand integration. The accelerating progression of digital technologies, which has evolved faster than theunderlying processes used to deploy, extendand manage om our perspective, four key activities defineDevOps (see Figure 2). Further, the role oftesting in DevOps can be best understood bycomparing how testing is approached in Agilevs. DevOps, as illustrated in Figure 3 (next page).Figure 1DevOps at a GlanceCollaborative Development:Increased collaboration between teamsPlan andMeasureContinuous Integration & Continuous Testing:Integration of software testing with deployment and operationsMonitor andOptimizeContinuous Release and Deployment:Release andDeployIncreased delivery speed and frequencyContinuous Monitoring:Improved quality by monitoring production performanceFigure 2cognizant 20-20 insightsDevelop andTest2

DevOps vs. AgileTesting in AgileTesting in DevOpsTest as early and as oftenas possibleTest continuouslyAutomate testing as muchas possibleAutomate almosteverythingContinuous integration and Continuous integration andtesting is a step forwardtesting is mandatoryPotentially shippable codeat the end of a sprintPotentially shippable codefollowing every integrationAs illustrated in Figure 5, the three main focalpoints of DevOps practices are automated buildand testing, configuration management, andcontinuous integration/continuous deployment.As a result, a new software development lifecycle(SDLC) is emerging, according to ForresterResearch (see Figure 6, next page). According toa Forrester4 survey, the characteristics of the newSDLC include: Faster release cycles: More than 24% ofcompanies have daily and weekly software releases. Changing requirements and lack of coordination between groups: Roughly 47% believethese disconnects were primary points of failure during a release review. The emergence of DevOps: Approximately66% of CIOs have this on their IT agenda.Figure 3DevOps: An Industry PerspectiveAccording to a recent industry report publishedby online educational resources purveyorDevOps.com,3 over 70% of IT organizations haveembraced DevOps in some form. The rush toDevOps is on, and for some, the journey has justbegun (see Figure 4).DevOps Adoption 2015Have few DevOps-powered projects30%Have DevOps-like initiatives in progress26%Have adopted DevOps24%Have not adopted DevOps but plan toHave no plans to adopt DevOps10%9%Source: DevOps.com and ElasticBox, d/.Response base: 300 participants, including DevOps specialists, IT engineers and executivesacross a wide range of organizations.Figure 4DevOps at a GlanceAutomated build & testing68%64%Configuration managementCI/CD63%62%Monitoring and metrics54%Collaboration between Dev & Ops53%Infrastructure as code49%Cloud application management41%Security & compliance36%Containerization29%Microservice managementSource: DevOps.com and ElasticBox, d/.Response base: 300 participants, including DevOps specialists, IT engineers and executivesacross a wide range of organizations.Figure 53

A New Software DevelopmentLifecycle Is EmergingOld SDLCNew cialized rolesEmpowered teamsResistant to changeOptimally responsiveOutsourcedAutomatedProject optimizationPortfolio optimizationSource: Forrester ResearchFigure 6Adoption ChallengesOrganizations born in the cloud era (aka, digitalnatives) adopt DevOps from day one; as such,DevOps practices match more organically withtheir DNA. DevOps is also gaining momentumwith legacy organizations but at a slower pacedue to a variety of challenges.The reasons are manifold. For starters, applications that large companies depend on typicallycontain extremely complex business logic;many have been in use for more than a decadeand have been updated and extended with newmission-critical functionality. Additionally, thereare development, operations and testing teamprocesses and practices, as well as the organization’s expectations that revolve around them.The Agile movement, for instance, has radicallychanged how applications are built and deployedby more closely aligning IT with the business,while operations and QA remain in silos.Typical challenges in IT organizations seeking toembrace DevOps include: Traditional Ops: This group spends the majority of its time putting out production fires, dealing with recurring failures and making urgentfixes. Test environment overhead: Each environment (testing, development, pre-production/staging and production) is unique. Operationsmust have a laser focus on production environment issues and, as such, does not have timeto work on improving infrastructure. cies without service virtualization techniquescause testing delays. A huge technical debt is accumulating in theseIT disciplines,5 and collaboration is often lackingbetween IT teams. The solution is to build DevOpspractices incrementally, one step at a time.Continuous Testing: A Step in theRight DirectionIn our view, continuous testing (CT) is the firststep in the right direction when embarking on aDevOps journey. CT is a metaphor for a continuousfeedback mechanism that drives software deliverythrough the SDLC tunnel. Automated feedback ateach checkpoint is an auto-trigger for the nextprocess in the delivery chain if the feedback isto move forward, or green. If the feedback is tonot move forward, the process immediately isstopped, and corrective measures are taken.Traditional IT organizations can shorten thepath to implement CT by reusing and realigningexisting test automation capabilities (as revealedin Figure 7, next page).Establishing a CT ecosystem involves the followingsteps at a high level: Shifting automation test scripts to an enterprise version control tool to establish asingle source of truth. Testing organizationshave commonly stored automation scriptsin test management tools or a shared folderstructure. However, an automation code base,just like an application code base, should reside in a version control repository. Integrating the automation suite with abuild deployment tool to enable centralizedexecution and reporting. Classifying the automation suite in multiplelayers of tests to enable faster feedback ateach checkpoint. Typical tests include:Traditional testing: Several thousand interdependent regression test cases run throughthe application user interface for severaldays, causing longer test cycles and manualhandoffs, resulting in much slower feedback.A loosely defined test data management strategy and downstream/upstream dependen-cognizant 20-20 insightsTraditional development: Late code integration results in integration issues and late defect fixes, leading to longer cycles. Scheduledelays often occur due to the absence of automated build verification and deployment procedures. Health check: An automated check verifiesservices are up after deployment. Typically,health checks run for just a few minutes. Smoketests: This most critical set ofautomated tests ensures that key systemfeatures are operational and no blocking4