X

Move your VMware and KVM applications to the cloud without making any changes

  • February 5, 2016

OPNFV Testing on Cloud

Author:
Brian Castelli
Brian Castelli is a software developer with Spirent creating test methodologies for today's networks. His current focus is on SDN and NFV.

The OPNFV project is dedicated to delivering a standard reference architecture for the deployment of carrier-grade Network Function Virtualization (NFV) environments. Testing is critical to the success of the project and to the success of real-world deployments, as evidenced by the many test-related sub-projects of OPNFV. One of those subprojects, VSPERF, is dedicated to benchmarking one of the key NFV components: The virtual switch.

The VSPERF community has developed a test harness that is integrated with several test tools. When deployed, virtual switches can be tested in a stand-alone, bare-metal environment. Standard benchmarking tests, such as RFC2544, are supported today, with more tests in the pipeline.

Running VSPERF for optimum performance and consistency requires dedicated hardware. This is acceptable for running the tests themselves, but it increases the cost of test development. Developers need a lower-cost environment where they can rapidly create and prove the functionality of tests that can then be moved to hardware for final testing. The hardware requirement also poses a problem for product marketing and sales. Those teams need a way to demo VSPERF test capabilities to potential customers without lugging around hardware.

The Ravello Systems environment presents a solution to the problems faced by these two groups. By virtualizing the test environment and taking advantage of Ravello’s Blueprint support, we were able to:

  • Virtualize the entire VSPERF test environment
  • Create a blueprint to enable rapid deployment
  • Create multiple, low-cost, access-anywhere development environments for engineers in various geographic locations
  • Create on-demand demo environments for customer visits and trade shows
  • Only pay for what we use

We started with virtual versions of our standard products.
 

  • A License Server is required to handle licensing of test ports
  • A Lab Server is required to handle REST API communication from the VSPERF test harness
  • STCv is our virtualized test ports. In this case, we used a single STCv instance with two test ports. This configuration gives our virtual test ports the best performance and consistency.

The 10.0.0.0 subnet was created for management. 11.0.0.0 and 12.0.0.0 were created for dataplane traffic.

Next we instantiated the VSPERF test harness host, connecting it to the appropriate networks.

All along the way, Ravello’s interface gave us the options we needed to configure.

  • DNS names. Each node was given a fully-qualified domain name that remained constant even though underlying IP addresses might change from one power-up to the next. This gave us the ability to script and configure without worry about changes.
  • CPU, memory, storage. Each node could have as much or as little resources as necessary. How many times have we been frustrated in the lab to find out that we need more cores or a larger hard drive? This is a configuration change instead of a purchase order.
  • Publish optimizes for cost or performance. We can minimize cost or publish to get better performance.
  • Timeout. By default, applications will timeout after a configurable period of time. This saves us from unnecessarily wracking up charges by accidentally running over the weekend. Of course, Ravello also supports “forever” operation for nodes that we really do want to run all weekend.

Overall, Ravello Systems is good value and flexibility for the needs described here.

Brian’s current role is to support the development of NFV test methodologies and to support Spirent’s participation in the OPNFV project.

Be the first to comment

Comments ( 0 )
Please enter your name.Please provide a valid email address.Please enter a comment.CAPTCHA challenge response provided was incorrect. Please try again.