At Open Networking Summit in March, we shared a progress report of our software-defined network build and introduced ECOMP, the brains of that network. To educate the industry and solicit feedback about this platform, we released a whitepaper that goes in depth to explain the purpose and parts of ECOMP. One of these components is Active and Available Inventory (A&AI), which maintains order and prevents chaos in this software-centric model for network infrastructure and services.
I could dive into the details, but first, let me paint a picture for you. Imagine you’re home and wanted to bake a cake. Before you can even start baking, you need to check your pantry for the required ingredients. Now, what if you didn’t know what flavor cake you wanted to bake in advance? Chocolate, devil’s food or vanilla? It gets trickier if you decide at a moment’s notice what to bake. To accommodate all those unknowns, you would need a “smart” pantry aware at all times of the available ingredients and how much you have.
So, what do baking and pantries have to do with A&AI? Well, A&AI is ECOMP’s “smart,” or dynamic pantry. It keeps track of what compute, network and storage resources are available, and which resources have been assigned and configured to provide services for our customers. When a customer orders a new service, ECOMP checks with A&AI (or looks in the pantry) to make sure we have all the resources to build the service. If we do, the MSO component of ECOMP executes the recipe (we really do call it that) to assign those resources and implement the customer’s order. We try to keep the network pantry full by monitoring what resources are assigned and available, and installing more equipment before we run out.
In addition to A&AI’s role to keep and track our compute, network and storage resources, it is also an inventory system. Network service providers have had inventory systems for decades, so what is new about ECOMP and why does it require a new approach? The answer lies in the network and service inventory changes needed by AT&T’s software-centric services. These new requirements are the reduced time scale, distribution of network functions globally and need to maintain accurate inventory data.
First, let’s first look at the reduced time scale. When a customer places an order for a software-centric network service, our goal is to deliver their service in minutes – not weeks. As resources get assigned, the inventory must be updated in real time, so we know what remains in the pantry. We have to support new services without having to develop, test and deploy new software. A&AI supports an approach driven by a metadata model, where we can add new resources and services by simply updating metadata, and without writing code. This model driven architecture is a part of all ECOMP components, not just A&AI.
Second, in terms of distribution, ECOMP’s distributed controllers assign resources locally, but those assignments must be rolled up to a centralized data store to reflect a customer’s global service connectivity. Finally, accuracy and timeliness are essential, since inventory errors will cause customer orders to fail. Scaling and synchronization challenges appear because of the distributed nature of network services covering resources in thousands of locations. These challenges don’t emerge when cloud services are provided in a handful of centralized datacenters. However, as the centralized data store of all network and customer data, A&AI enables near real-time automated trouble reporting and resolution processes.
A&AI is ECOMP’s dynamic pantry that ensures we always have the right ingredients to make what our customers order.
Sal Talamo, Assistant Vice President, AT&T ECOMP