You don’t say your car is broken when you get a flat tire. Sure, it’s not road-ready at that moment. But you’ve pinpointed what needs to be done to get the car back in working order.
Treat your SDN environment similarly. Perhaps some workloads in your datacenter aren’t performing as expected, but the application still works, and overall your users are not being affected. You still want to see what specific parts of the infrastructure (physical or virtual) are affecting the application, and assess what is the impact on business.
The first step in establishing a solid SDN operation – one that delivers a great and consistent quality of experience – is having visibility into how everything in this environment performs together, from the bare metal all the way to the application itself.
Data mediation is the start – it’s being able to translate data from different elements, each with its own representation of the infrastructure into a simple and consistent structure that you can integrate and remix with other systems and tools to allow you to visualize and analyze your environment in the right way.
In order to properly analyze performance in SDN deployments, you need to think like a doctor diagnosing a patient. When a problem arises, start with the broadest questions. For example, assess the degree of degradation of an application response time, and then narrow down until you’ve eliminated the majority of options and honed in on one or two potential causes. Perhaps a database container is overloaded or a load balancing NFV is misconfigured.
By placing importance on analyzing data generated from your environment from the bare metal all the way to your VMs and applications, you’ll also be able to find patterns during forensic analysis and spot performance trends over time. When you launch a new application, you can’t predict how it will evolve based on what you know from static infrastructure, but if you have the data, it can help you understand how customers interact with your application and how the infrastructure impacts performance as it evolves over time.
You can start by analyzing, in near-real time, a variable like CPU utilization and notice when it hits a certain percentage. And, once you’ve noted the trends and patterns, you can ascertain whether or not that’s normal behavior for your application by looking at other variables that are related to what defines your application in the infrastructure sense. And remember, “normal” will change over time and you can’t think of it as a static metric.
Using that logic, if that percentage of my CPU is being utilized and both my application page serve time and the network load is increasing, I may likely have a problem. But if that same percentage of my CPU is being utilized and network load has increased, but page serve time is remaining stable, I do not have a critical issue. It means I am using more of my infrastructure to its capacity.
Analysis like this helps enterprises make better business decisions. Analyzing your SDN data is not only about catching problems before they happen. It is about optimizing your infrastructure so you can deliver the best quality of experience to customers while managing your investment.
With solid analytics, you can make short-term operational decisions that allow you to keep your applications running well and your customers happy, while also getting a starting point towards highlighting where you can optimize your infrastructure to better align to your business goals.