Updated: June 2015
What is OpenConfig?
OpenConfig is a collaborative effort by network operators to develop programmatic interfaces and tools for managing networks in a more dynamic, vendor-neutral way.
What problem is OpenConfig addressing?
As network operators seek to adopt a more dynamic, programmable infrastructure, the lack of a common way to configure and manage network elements is becoming a significant source of increased complexity, poor reliability, and greater cost. Key focus areas being addressed by OpenConfig include development of vendor-neutral APIs for declarative configuration, modern, scalable monitoring solutions, and adoption and native support on vendor platforms.
What is OpenConfig producing?
OpenConfig’s initial focus is on compiling a consistent set of vendor-neutral data models (written in YANG) based on actual operational needs from use cases and requirements from multiple network operators. These models may be developed directly by OpenConfig, or compiled from 3rd party modules that conform with the OpenConfig requirements. OpenConfig is interacting with standards bodies, and network equipment manufacturers - and intends that these models will become the basis of widely-adopted, standardized interfaces.
OpenConfig also aims to produce guidelines for a consistent approach to data modeling - encompassing both configuration and operational state - which is uniformly supported across network vendors. The OpenConfig models aim to support real-world use cases based on feedback from the operators that contribute to the effort.
Finally, OpenConfig participants are also publishing open source tools for manipulating and producing code artifacts from YANG models.
Who are the members of OpenConfig? What is the governance model?
OpenConfig is a working group started by technical contributors from Google, AT&T, British Telecom, and Microsoft. There is no formal governance at present – we depend on participants having shared vision and goals, behaving transparently, and supporting rough consensus. Having multiple operators with different kinds of deployment and operational considerations ensures that a broad set of use cases, requirements, and vendor experiences are incorporated in OpenConfig. Other interested network operators who share OpenConfig’s goals and vision are welcome to join the effort (see the Operator FAQ).
Vendors are already working on data models – why not adopt those?
We believe that these data models should be vendor-neutral and based on the needs and practices of network operators, and that operators should play an active role in developing them. In particular, OpenConfig represents a user / customer perspective that is focused on meeting real operational needs. Nevertheless, we do work in partnership with vendors to ensure OpenConfig models can be supported natively on various platforms.
How can vendors contribute to help this effort?
Vendors can participate in OpenConfig, and provide feedback on models, through mailing list discussions, and via standards channels (e.g., for models published in the IETF). Most importantly, we encourage networking vendors to support the models for configuration and operational state natively on their platforms.
How does this relate to similar efforts in standards bodies like IETF?
OpenConfig is not a standards organization, but we hope to complement standards efforts, including those in IETF. OpenConfig contributors are active in several relevant IETF working groups. We will continue to publish models in IETF or other communities, as appropriate, to provide an operator / customer perspective on how data models should be built to enable more programmable networks.
Will OpenConfig models be standardized?
As we work with vendors to implement support for the data models, we expect that some or all of them would be the basis for standard models produced in the IETF or other organizations. Standardization, while important for interoperability, is a relatively lengthy process. We see a lot of value in working with vendors on implementations to gain experience and insights that would inform and improve any subsequent standard models.
Isn’t there a risk that vendors will have to implement multiple models?
OpenConfig is engaged directly with both vendors and standards organizations on developing consistent, operationally-focused data models that are easy to program. We strive for agreement on the operational requirements to encourage convergence across models in terms of their structure, usability, relevance, and content. In cases where standards organizations are not able to produce models that meet operational requirements, OpenConfig will develop models that we need, and intends to work directly with vendors on implementations. By consolidating requirements across a significant number of operators, we hope that guidelines and requirements produced by OpenConfig will be carefully considered.
Does OpenConfig plan to adopt or endorse a specific transport protocol or data encoding scheme for sending and receiving configuration and state data?
Not at this time. Our primary focus is on driving adoption and support of vendor neutral data models, i.e., the structure, content, and semantics of the data. There are a number of possible transport and encoding options for the data, and operators may have different preferences based on factors such as maturity of implementations, integration with existing management stacks, and availability of tooling.
Which vendors are implementing support for OpenConfig models?
At this time, several vendors have in-progress implementations at different stages of maturity. The current implementation focus is on the OpenConfig BGP and routing policy models, which are the most mature and have gone through the most review and iteration. The specific vendors will be announced at a future time.