< back to overview

Intent: What. Not How.

Jan 12, 2015
Sue Kim - gu
Sue Kim - gu About the author

HP’s Dave Lenrow discusses the general approach of intent-driven networking and how it relates to ONF’s NBI Working Group.

What if, instead of describing how we want a network to be configured, we could describe how distributed workloads behave? Instead of requiring detailed network configuration expertise and understanding network protocols and equipment interfaces, network managers would only need to specify the distributed workload’s behaviors and communications requirements.

For the purposes of this conversation, let’s call the “what I need” description intent and the “do it like this” description prescription. Sometimes this approach is conflated with the much broader category of “policy-based” infrastructure. Intent can be considered an aspect of a policy-based approach, yet not all policy-based approaches are based on intent.

Intent-driven systems shift the focus from networking details to distributed network application. This will make network programmability far more accessible, as opposed to the far smaller and specialized group of network experts.

Intent is “what” not “how”
When one is asked to build a network to support new sales automation software, the high level goal does not involve routing or switching, but is to enhance sales automation. Routing is not intent. MPLS is not intent. Layer 1 (or 2 or 3…) is not intent. Vendor-specific feature X is not intent.

However, any of these networking solutions might satisfy the business intent. You can get more precise without getting prescriptive. Low latency is intent. High speed is intent. Twenty megabits per second is intent. Highly secure is intent. Best possible transport quality is intent.

Is gold-class service intent? No! “Why won’t bronze-class work?” “Because the application can’t tolerate interference from bursty, noisy neighbors.” Aha! Intent. Applications want isolation, or constant bitrate or an SLA, which might be implemented with gold-service by one provider and by a dedicated optical wavelength by another. Requesting two aspirin is also not intent. It is not “portable” because it is subject to “Sorry, I’m out of aspirin.” Make my headache go away is intent, which can be satisfied by ibuprofen, Tylenol, morphine, hydration, meditation, sleep, etc. Intent is “Please cut my hair,” and prescription is “Here is a list of all my hairs and the length I want each one cut to.” Intent is “connect server A to server B.” Prescription is “create a new OSPF route on switch 27 and update ACL lists on the following switches…”

Intent in ONF NBI 
Intent is compelling because of the ease of use benefits and the portability benefits for network operators. An intent-based network controller NBI has several additional benefits from an architectural perspective.

For all of these reasons, the ONF NBI Working Group is working on a project to specify the architecture and initial information model for an intent-based interface to the SDN controller. We’ll ultimately build two or more reference implementations of this NBI on different SDN controller platforms.

ONF agrees that the time is right to leverage the diverse research, thinking, and early development in this area and propose a use-case-driven common interface definition with open source implementations before the world diverges on multiple proprietary, vendor-specific approaches. This is a critical interface between cloud orchestration systems (e.g., OpenStack) and SDN controllers. An intent solution will support the kind of portability, composability, and scalability demanded by the cloud.

A diverse team of operators and vendors are to collaborate on this project, with input from the open source community as well. More information about this work and our meetings can be found here. This is a great opportunity to make important contributions to the future of infrastructure, and we invite one and all to participate.

– Dave Lenrow, Distinguished Architect, Advanced Technology Group – HP Networking

David Lenrow is an architect in the CTO’s office at Hewlett Packard Networking, TSC member in OPNFV and OpenDaylight, vice-chair cloud/data center in the ONF NBI Working Group, and active in OpenStack and OCP communities. He was trained as a computer scientist and has spent more than 20 years driving innovation in digital technology with an emphasis on networks, storage, and media. He is passionate about open innovation and seeking to make contributions to the cloud revolution currently sweeping the tech industry. @dlenrow, david.lenrow@hp.com

Sue Kim - gu