Battle test your virtual device with impairments

NFV is used in the datacenter, service provider and any physical server. However, before replacing the core of the infrastructure or any part of it , the virtual network function must be completely characterized. In addition to functional and performance tests, impairment tests are a crucial piece of the puzzle.

Most public tests for NFV focus on flexibility and scale, which are the strengths of virtualization.

Everyone is waiting for the adoption to kick in because of the fears of the unknown. We need to allay fears of using NFV by showing realistic test scenarios. Failure or impairment of links is very common occurrence is the real world. There are several models of failure at the packet, link and network level. The failure modes of physical links could be link flap, pkt drops, reordering, delay etc.  Virtualization adds more unpredictability because of shared nature of the infrastructure.  Do you now feel confident enough to replace the core of your ISP?

I recently found a great proof-of-concept demo by OnLab in ONS. This was part of the CORD project (CORD stands for Central Office ReArchitected as a Datacenter). In their demo, their vCPE was a VM running on a server, with the intention of making one vCPE per user/subscriber. This is a very interesting POC, with the stated intention of “Re-architecting the Central Office”. The end goal is to make very cheap and dumb devices at the customer premises, and move all the intelligence to massive datacenters. This gives the ability to great control, flexibility and scale.

NFV is used in the datacenter, service provider and virtually any physical box with anything to do with networking. However, before a service provider replaces the core of the infrastructure or any part of the infrastructure, the NFV must be completely characterized. In addition to the functional test, performance test, impairment test is a crucial piece of the puzzle.

Another example is a Hadoop cluster, a massively parallel operation that runs on a cloud. A Hadoop clusters are commonly referred as “shared nothing” systems, because the only thing shared between the nodes is the network. How would a congested VNF adding delay impact the performance of the entire cluster? 

Testing Virtual Devices with Impairments

Prepare for real-life failure scenarios in your test bed

Imagine being able to run these real-life failure scenarios in your test bed and make your VNF battle ready? What’s needed is a way to add impairments to an otherwise healthy link to be able to test these scenarios. A solution such as the Spirent Attero Virtual can help you add these impairments to your NFV test methodology to help you characterize the performance and behavior of your VNF. It can help bring the impairment functionality to the virtual world. Spirent Attero virtual is a virtual avatar of the same and  has five easy steps for creating impairments – Connect, setup interface, select Flows, add impairments, and view stats Spirent Attero can quickly add, bandwidth impairments (packet loss), delay/jitter and bit errors (packet corruption).

To help service providers and cloud service providers transition to NFV, we have to provide them the data to prove the maturity of the solutions. I am eagerly waiting to see the transition from the POC stage which is where we are now to a full blown implementation.

To learn more about adding impairments to validate your virtual network. 

comments powered by Disqus
× Spirent.com uses cookies to enhance and streamline your experience. By continuing to browse our site, you are agreeing to the use of cookies.