Welcome!

@DevOpsSummit Authors: Yeshim Deniz, Zakia Bouachraoui, Pat Romanski, Liz McMillan, Elizabeth White

Related Topics: @DevOpsSummit, Microservices Expo, Linux Containers, Containers Expo Blog, @CloudExpo

@DevOpsSummit: Blog Post

What Does an SLA Mean to You? | @DevOpsSummit #DevOps #SLA #ContinuousDelivery

You can factor performance and load testing into your continuous integration process by focusing on performance SLAs

When it comes to testing an application, many project managers and test leads do not fit performance and load testing early on in the development life cycle. Often, performance and load testing are conducted once the application has been completed and after all functional testing is done. In fact, it's frequently a last step - almost an afterthought - before the app is ready to go into production.

The problem with this approach is the classic problem with late-stage testing. When issues are found by the testers, developers have to go back and open up long-finalized code to fix them. There's risk in causing other parts of the application to break. Addressing problems after-the-fact becomes time consuming and expensive. Furthermore, any delay in releasing a new feature or a new app can directly impact revenue.

Even though you are running an agile development process, that doesn't necessarily mean that your performance testing is being conducted in a truly agile way. Saving performance testing for a "final sprint" before release still treats it like a waterfall development step, with all the cost and risk that comes with that. In this post, we will show you how to make load testing happen early and often by putting SLAs on the agile task board.

What Does an SLA Mean to You?
You can factor performance and load testing into your continuous integration process by focusing on performance SLAs, or service level agreements. Traditionally, SLAs are contracts put in place between two parties, such as a service firm and its client. This agreement sets various attributes about how the service will be delivered. For example, a large company that outsources its help desk operations to another company may require that all incoming emails be responded to within one hour. That agreement allows the client to build its business knowing that the desired service will be executed according to its specifications.

As a performance tester, you can think of an SLA in terms of your performance parameters. In your case, it's an implicit contract that you make with your user. It's almost like a standard that you hold your application to on behalf of your users. Your SLA may state that all pages need to load within 2 seconds, or that the first page of results of a search should be displayed within 4 seconds.

Though this isn't a hard-and-fast contract, that's how you want to think about it. Specify your SLA items during the application design process, and you can turn them into requirements that are placed on the task board as part of agile. That way, developers are taking performance into account as they develop the application, and not leaving it as an afterthought.

What Your SLA Should Include
There are all sorts of website metrics you can include in an SLA. The most important thing is to get something specified up front. This is how you are going to get developers coding for performance at the very beginning of the process.

When you are first starting, we recommend that you have enough specified in your performance SLAs to support automated smoke testing. Consider what happens in a smoke test. Once the application is built, it will be deployed and a series of tests will be run to make sure that everything works. This basic test might be as simple as setting up a user and having the user log in. Adding a performance SLA to that smoke test could be as simple as requiring that the login process be completed within a given time window.

Add that requirement to your task board, and suddenly you are automating performance testingright at the beginning of the development process. As more of the app gets developed, you can add more requirements alongside the modules and features being created.

When you create your SLAs, it's a good idea to identify a set of key pages to test regularly (home page, shopping cart, contact form, search results, chat window, etc.), and define requirements for a set of metrics such as:

  • DNS resolution time
  • Time-to-display
  • Time-to-last-byte
  • System uptime

Obviously your application will have its own critical performance metrics as well.

Finally, define a base level of network types (Wi-Fi, 3G) and platforms (operating systems and browsers) that your SLA will be measured from. Requiring zippy load times for someone running IE6 over a dial-up internet connection will only prompt hatred and ire from your developers, but providing realistic expectations that apply to the bulk of your users will keep everyone happy.

How Often to Conduct Load and Performance Testing
You now know how set expectations with SLAs in place, but how often should performance and load tests to be conducted? Remember, testing is happening throughout the agile development process, and at different scales. For many agile shops, your automated build process includes a number of tests that get executed every time a build happens. Then you may have additional testing that is done on a more thorough basis, but less frequently. Sometimes you have test scripts that are run every time code is checked in.

Match your automated performance testing to the process. It's unreasonable to expect a full load test be conducted every time the app is smoke tested, but something simple will match the scope, intent, and turn-around time of the smoke test, fitting in nicely with what's already happening. Use your best judgment and align your SLAs with the work being done, and you should be ok.

The SLA: One Secret to Quick and Efficient Load Testing
As your application grows and your process matures, you'll end up with a rich library of SLA requirements throughout your portfolio of test structures. You'll also have automated tests that can validate those SLAs within the right testing constructs: smoke testing, comprehensive automated tests, unit tests, and other areas. It won't take too many cycles for performance to become a regular expectation of the code that gets produced.

Photo Credit: Enrique Fernández

More Stories By Tim Hinds

Tim Hinds is the Product Marketing Manager for NeoLoad at Neotys. He has a background in Agile software development, Scrum, Kanban, Continuous Integration, Continuous Delivery, and Continuous Testing practices.

Previously, Tim was Product Marketing Manager at AccuRev, a company acquired by Micro Focus, where he worked with software configuration management, issue tracking, Agile project management, continuous integration, workflow automation, and distributed version control systems.

@DevOpsSummit Stories
Dion Hinchcliffe is an internationally recognized digital expert, bestselling book author, frequent keynote speaker, analyst, futurist, and transformation expert based in Washington, DC. He is currently Chief Strategy Officer at the industry-leading digital strategy and online community solutions firm, 7Summits.
Addteq is a leader in providing business solutions to Enterprise clients. Addteq has been in the business for more than 10 years. Through the use of DevOps automation, Addteq strives on creating innovative solutions to solve business processes. Clients depend on Addteq to modernize the software delivery process by providing Atlassian solutions, create custom add-ons, conduct training, offer hosting, perform DevOps services, and provide overall support services.
Contino is a global technical consultancy that helps highly-regulated enterprises transform faster, modernizing their way of working through DevOps and cloud computing. They focus on building capability and assisting our clients to in-source strategic technology capability so they get to market quickly and build their own innovation engine.
The standardization of container runtimes and images has sparked the creation of an almost overwhelming number of new open source projects that build on and otherwise work with these specifications. Of course, there's Kubernetes, which orchestrates and manages collections of containers. It was one of the first and best-known examples of projects that make containers truly useful for production use. However, more recently, the container ecosystem has truly exploded. A service mesh like Istio addresses many of the challenges faced by developers and operators as monolithic applications transition towards a distributed microservice architecture. A tracing tool like Jaeger analyzes what's happening as a transaction moves through a distributed system. Monitoring software like Prometheus captures time-series events for real-time alerting and other uses. Grafeas and Kritis provide security polic...
DevOpsSUMMIT at CloudEXPO will expand the DevOps community, enable a wide sharing of knowledge, and educate delegates and technology providers alike. Recent research has shown that DevOps dramatically reduces development time, the amount of enterprise IT professionals put out fires, and support time generally. Time spent on infrastructure development is significantly increased, and DevOps practitioners report more software releases and higher quality. Sponsors of DevOpsSUMMIT at CloudEXPO will benefit from unmatched branding, profile building and lead generation opportunities.