Crafting the Test Approach
This chapter covers the following topics:
- Motivations for Different Types of Testing
- Test Scoping
- Test Planning
- Choosing the Right Test Tools
- Writing the Test Plan
Chapter 1, "A Business Case for Enterprise Network Testing," stressed the importance of assessing the business reasons for testing as your first step in crafting an effective test approach. In the same way that a network designer would be foolish to specify equipment or make technical recommendations without prior knowledge of customer requirements, a test engineer would be misguided to attempt writing a test plan without first understanding the triggers, scope, motives, and expectations for the test initiative. By rushing ahead and skipping this critical step, you risk missing the mark in your testing, focusing on the wrong types of tests, or capturing erroneous results. This will waste precious time and resources as you continuously redefine your test plan; add, remove, or modify equipment to your lab topology; rerun your test cases; and generate reports. Taking time to identify the objectives and outline an assessment is critical before you ever step foot into the lab. Only after the following questions are answered should you begin to write a detailed test plan or build a lab topology:
- What are the test triggers?
- Who is requesting the test and what are their motives?
- How much testing is necessary and what constitutes success?
- What is the impact of test failure and what are the known risks?
- What are the resources (people, lab equipment, and test tools) required to execute the test?
As discussed in Chapter 2, "Testing Throughout the Network Lifecycle," a complimentary relationship between network testing and design functions exists in organizations that execute enterprise architecture effectively. We explained how structured testing complements and validates design deliverables, by providing examples of the different types of test requests that you can expect throughout the network's lifecycle.
This chapter will begin to fill in the practical details of what is necessary to build an effective approach toward different types of test requests. It begins with a suggested approach for assessing and scoping a test project, and offers guidance and best practices for the following considerations:
- How to identify test case scenarios
- How to develop a lab prototype
- How to choose the proper test tools necessary to execute the different types of tests
- How to write a detailed test plan
As with most technical undertakings, there is no absolute right way to approach systems testing. We do not promote ours as the only way to conduct successful testing. However, this is a proven method that will improve your chances of getting it right the first time.
Motivations for Different Types of Testing
The first step in assessing the objective and scope of a test effort is to understand the reasons for why it was requested, and the motives of the people or organization that requested it. In some instances, your client may be able to clearly tell you why they want testing and what they expect from testing, while others may only be able to tell you that their proposed deployment "is critical to the business and must be tested." In cases of the latter, you will need to rely on knowledge of your client, personal experience, and industry best practices to determine the objective and scope of the test effort. Following are some of the most common triggers and motivations associated with the different types of testing.
Proof of Concept Testing
Proof of concept (POC) testing is normally conducted during the Plan Phase of a new network design, or prior to the introduction of a new technology or service into an operational network. A network architect will often request that a POC test be completed to ensure that a new product or technology will work as expected in the context of their design. Successful POC testing is often the criteria for purchasing or moving into the low-level design (LLD) phase of a project, and in some cases POC testing is a mandatory milestone to be completed before purchasing approval will be granted. In general, POC testing should be conducted systematically but persist only as long as necessary to prove that a proposed solution will work as expected. An exception to this general rule is when POC testing is used as a means to differentiate between similar products as part of a "bake-off" test. These types of tests often require extensive scale and feature testing in order to provide the necessary data to differentiate between competing products.
Network Readiness Testing
Network readiness testing is often included as part of a network assessment to determine whether a production network can meet the needs of a new application or service, and to identify any gaps that may hinder it. This type of testing is commonly conducted prior to deploying a Cisco Unified Communications (UC) solution, to help an enterprise determine whether its network will be able to meet the stringent requirements associated with real-time applications. Network readiness testing for UC often involves test tool injection and measurement of synthetic application traffic across a live network to predict how the actual application will perform when network elements are running in steady-state conditions, during day-to-day operations. Success criteria for this type of testing is easy to define because the SLA requirements with respect to delay, jitter, and loss are well understood for UC applications. Careful planning and coordination is often necessary when this type of network readiness testing is conducted so that production service disruption can be avoided.
Design Verification Testing
As the name suggests, this type of testing occurs during the Design Phase of a network's lifecycle. Design verification testing is similar to POC testing in that both are performed in order to gain confidence in a proposed design or solution prior to deployment. Design verification testing is typically more extensive than POC testing, however, as it often represents the last opportunity before implementation to fully examine whether all aspects of a design will function as expected when subjected to various stress conditions. Design verification testing is focused on performance, scalability, failover, operations, and manageability elements of a design. The output from this type of testing often feeds into the software recommendations, hardware specifications, and device configuration templates of an LLD document.
Hardware Certification Testing
Hardware certification testing often occurs during the Optimize Phase of a network's lifecycle as new platforms are introduced into existing operational networks to provide enhanced capabilities, better performance, or to replace equipment that is reaching end-of-life (EOL) status from a vendor supportability standpoint. Engineering and operations groups of an enterprise often require that hardware certification testing be completed before a product can be deployed in the production network. While it is generally accepted that equipment vendors will subject new platforms to a variety of tests during the product development cycle, there is no substitute for customized, enterprise-specific testing to uncover defects or feature limitations that would not be found otherwise. It is nearly impossible for an equipment manufacturer to predict how a customer might deploy every feature, or the level of stress that a platform might be subjected to in an operational network with unique requirements. Likewise, it would be impractical for an equipment vendor to perform interoperability testing with every other vendor's equipment that might be deployed on a customer network. Hardware certification tests generally are simple in nature and shorter in duration as compared to other tests because they focus mainly on "unit level" test cases that can be conducted on relatively small lab topologies.
Network Operating System Testing
This type of testing is often required by the operations teams responsible for OS upgrades and is similar in scope to hardware certification testing. Network OS testing is often performed during the Optimize Phase of a network's lifecycle, as operating software reaches its end of life, or when new features or bug fixes are needed. Overall, there are many different levels of network OS testing that can be undertaken, some of which are only appropriate during the product development phase by the equipment vendor test groups. The most common types of tests conducted by clients are software acceptance tests, which are a customized suite of tests executed to verify general feature functionality. Regression tests are a variant of software acceptance tests, in which critical features that worked in the past are retested to ensure that they are still functioning properly in the new OS. The scope of network OS testing ranges from small, short-duration tests (such as bug fix verifications), to longer-duration, multithreaded tests that involve multiple features to be verified in parallel.
Migration Plan Testing
One of the most challenging and critical aspects of a networking project is the migration of users and services to a new network infrastructure. Even the best network designs are destined for failure if they cannot be implemented without causing extended service outages. Yet despite the risks, many network architects spend a disproportionate amount of time focused on the "end state" of their network designs, developing migration plans as an afterthought, if at all. A good migration plan should address how routing protocols and applications will interact when the network is partially migrated, providing success indicators and a backout plan when unexpected behavior is encountered during a migration. Testing of a migration plan is an essential part of the design process for networking projects of any scale. It is sometimes a requirement of the implementation or operations groups responsible for making changes to the network. In some instances, a migration plan can be developed during a design verification lab test effort by repeating the baseline and performance test scripts on the interim topology consisting of the old and new networks.
A high-level migration test plan approach for a new network backbone might look something like this:
- Step 1. Build a prototype of the old and new network backbone topologies.
- Step 2. Run a baseline test using known traffic patterns from the existing and new networks.
- Step 3. Physically and logically interconnect the old and new network backbone topologies, as they will be connected during the migration. If this will take multiple steps, each interim topology should be tested.
- Step 4. Run the same set of baseline tests on the interim network that you ran on the old network.
- Step 5. Simulate device and circuit failure scenarios in each interim step of the migration in order to understand the impact on test traffic and whether any collateral damage occurs.
- Step 6. Disconnect the old portion of the network or reprovision it on the new backbone. This should be done the same way as the migration plan will be done. If this will be done in multiple steps in your plan, you should test each one of them.
- Step 7. Repeat the set of baseline tests.
- Step 8. Run a set of new tests that exercise any new features or services to be offered by the new network.
A migration test would be considered successful when the baseline test results meet or exceed the performance of the old network and the features offered by the new network are verified.
Network Ready for Use Testing
A network ready for use (NRFU) test typically is executed on a new greenfield network infrastructure as a last step in certifying that it is ready to carry production traffic. During an NRFU test, network devices are methodically checked to ensure that they have been implemented according to the design specifications and are running in an error-free state.
Some of the tests commonly associated with NRFU testing include the following:
- Device tests (hardware/software inventory, power, syslog error checking)
- Circuit tests (throughput, delay, jitter, errors)
- Routing tests (adjacencies, routing table consistency)
- Traffic tests (end-to-end traffic testing)
- Network service tests (multicast, QoS, WAN acceleration)
- Application tests
- Management/NMS/security tests
In some cases, the NRFU testing extends to a limited production pilot where a low-risk site or portion of the network is cut over to the new network and monitored closely for a "probationary" period of time.