Internet-Draft | Network Inventory YANG | October 2022 |
Yu, et al. | Expires 27 April 2023 | [Page] |
This document defines a YANG data model for network hardware inventory data information.¶
The YANG data model presented in this document is intended to be used as the basis toward a generic YANG data model for network hardware inventory data information which can be augmented, when required, with technology-specific (e.g., optical) inventory data, to be defined either in a future version of this document or in another document.¶
The YANG data model defined in this document conforms to the Network Management Datastore Architecture (NMDA).¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 27 April 2023.¶
Copyright (c) 2022 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
Network hardware inventory management is a key component in operators' OSS architectures.¶
Network inventory is a fundamental functionality in network management and was specified many years ago. Given the emerging of data models and their deployment in operator's management and control systems, the traditional function of inventory management is also requested to be defined as a data model.¶
Network inventory management and monitoring is a critical part for ensuring the network stays healthy, well-planned, and functioning in the operator's network. Network inventory management allows the operator to keep track of what physical network devices are staying in the network including relevant software and hardware versions.¶
The network inventory management also helps the operator to know when to acquire new assets and what is needed, or to decommission old or faulty ones, which can help to improve network performance and capacity planning.¶
In [I-D.ietf-teas-actn-poi-applicability] a gap was identified regarding the lack of a YANG data model that could be used at ACTN MPI interface level to report whole/partial network hardware inventory information available at domain controller level towards north-bound systems (e.g., MDSC or OSS layer).¶
[RFC8345] initial goal was to make possible the augmentation of the YANG data model with network inventory data model but this was never developed and the scope was kept limited to network topology data only.¶
It is key for operators to drive the industry towards the use of a standard YANG data model for network inventory data instead of using vendors proprietary APIs (e.g., REST API).¶
In the ACTN architecture, this would bring also clear benefits at MDSC level for packet over optical integration scenarios since this would enable the correlation of the inventory information with the links information reported in the network topology model.¶
The intention is to define a generic YANG data model that would be as much as possible technology agnostic (valid for IP, optical and microwave networks) and that could be augmented, when required, to include some technology-specific inventory details.¶
[RFC8348] defines a YANG data model for the management of the hardware on a single server and therefore it is more applicable to the domain controller South Bound Interface (SBI) towards the network elements rather than at the domain controller's northbound. However, the YANG data model defined in [RFC8348] has been used as a reference for defining the YANG network hardware inventory data model presented in this draft.¶
For optical network hardware inventory, the network inventory YANG data model should support the use cases (4a and 4b) and requirements as defined in [ONF_TR-547], in order to guarantee a seamless integration at MDSC/OSS/orchestration layers.¶
The proposed YANG data model has been analysed at the present stage to cover the requirements and use cases for Optical Network Hardware Inventory.¶
Being based on [RFC8348], this data model should be a good starting point toward a generic data model and applicable to any technology. However, further analysis of requirements and use cases is needed to extend the applicability of this YANG data model to other types of networks (IP and microwave) and to identify which aspects are generic and which aspects are technology-specific for optical networks.¶
This document defines one YANG module: ietf-network-inventory.yang (Section 4).¶
Note: review in future versions of this document the related modules, depending on the augmentation relationship.¶
The YANG data model defined in this document conforms to the Network Management Datastore Architecture [RFC8342].¶
The following terms are defined in [RFC7950] and are not redefined here:¶
The following terms are defined in [RFC6241] and are not redefined here:¶
The terminology for describing YANG data models is found in [RFC7950].¶
TBD: Recap the concept of chassis/slot/component/board/... in [TMF-MTOSI].¶
Following terms are used for the representation of the hierarchies in the network hardware inventory.¶
Network Element:¶
a device installed on one or several chassis and can afford some specific transmission function independently.¶
Rack:¶
a holder of the device and provides power supply for the device in it.¶
Chassis:¶
a holder of the device installation.¶
Slot:¶
a holder of the board.¶
Component:¶
holders and equipment of the network element, including chassis, slot, sub-slot, board and port.¶
Board/Card:¶
a pluggable equipment can be inserted into one or several slots/sub-slots and can afford a specific transmission function independently.¶
Port:¶
an interface on board¶
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
A simplified graphical representation of the data model is used in Section 3 of this document. The meaning of the symbols in this diagram is defined in [RFC8340].¶
In this document, names of data nodes and other data model objects are prefixed using the standard prefix associated with the corresponding YANG imported modules, as shown in the following table.¶
Prefix | Yang Module | Reference |
---|---|---|
ianahw | iana-hardware | [RFC8348] |
ni | ietf-network-inventory | RFC XXXX |
yang | ietf-yang-types | [RFC6991] |
RFC Editor Note: Please replace XXXX with the RFC number assigned to this document. Please remove this note.¶
Based on TMF classification in [TMF-MTOSI], inventory objects can be divided into two groups, holder group and equipment group. The holder group contains rack, chassis, slot, sub-slot while the equipment group contains network-element, board and port. With the requirement of GIS and on-demand domain controller selection raised, the equipment room becomes a new inventory object to be managed besides TMF classification.¶
Logically, the relationship between these inventory objects can be described by Figure 1 below:¶
In [RFC8348], rack, chassis, slot, sub-slot, board and port are defined as components of network elements with generic attributes.¶
Considering there are some special scenarios, there is no direct relationship between the rack and network element. In some cases, one network element contains multiple racks while in other cases one rack contains several shelves belonging to one or more network elements.¶
While [RFC8348] is used to manage the hardware of a single server (e.g., a network element), the Network Inventory YANG data model is used to retrieve the network hardware inventory information that a controller discovers from all the network elements under its control.¶
However, the YANG data model defined in [RFC8348] has been used as a reference for defining the YANG network inventory data model. This approach can simplify the implementation of this network hardware inventory model when the controller uses the YANG data model defined in [RFC8348] to retrieve the hardware from the network elements under its control.¶
Note: review in future versions of this document whether to re-use definitions from [RFC8348] or use schema-mount.¶
+--ro network-inventory +--ro equipment-rooms | +--ro equipment-room* [uuid] | +--ro uuid yang:uuid | ................................... | +--ro racks | +--ro rack* [uuid] | +--ro uuid yang:uuid | ................................... | +--ro contained-chassis* [ne-ref component-ref] | +--ro ne-ref? leafref | +--ro component-ref? leafref +--ro network-elements +--ro network-element* [uuid] +--ro uuid yang:uuid ................................... +--ro components +--ro component* [uuid] +--ro uuid yang:uuid ...................................¶
For all the inventory objects, there are some common attributes existing. Such as:¶
Identifier: here we suggest to use uuid format which is widely implemented with systems. It is guaranteed to be globally unique.¶
Name: name is a human-readable label information which could be used to present on GUI. This name is suggested to be provided by server.¶
Alias: alias is also a human-readable label information which could be modified by user. It could also be present on GUI instead of name.¶
Description: description is a human-readable information which could be also input by user. Description provides more detailed information to prompt users when performing maintenance operations.¶
Location: location is a common management requirement of operators. This location could be an absolute position (e.g. mailing address), or a relative position (e.g. port index). Different types of inventory objects may require different types of position.¶
module: ietf-network-inventory +--ro network-inventory +--ro equipment-rooms | +--ro equipment-room* [uuid] | +--ro uuid yang:uuid | +--ro name? string | +--ro description? string | +--ro alias? string | +--ro location? string | ................................... | +--ro racks | +--ro rack* [uuid] | +--ro uuid yang:uuid | +--ro name? string | +--ro description? string | +--ro alias? string | +--ro rack-location | | +--ro equipment-room-name? leafref | | +--ro row-number? uint32 | | +--ro column-number? uint32 | ................................... +--ro network-elements +--ro network-element* [uuid] +--ro uuid yang:uuid +--ro name? string +--ro description? string +--ro alias? string +--ro ne-location | +--ro equipment-room-name* leafref ................................... +--ro components +--ro component* [uuid] +--ro uuid yang:uuid +--ro name? string +--ro description? string +--ro alias? string +--ro location string ...................................¶
The YANG data model for network hardware inventory mainly follows the same approach of [RFC8348] and reports the network hardware inventory as a list of components with different types (e.g., chassis, module, port).¶
+--ro components +--ro component* [uuid] +--ro uuid yang:uuid +--ro name? string +--ro description? string +--ro class? identityref +--ro contained-child* -> ../uuid +--ro hardware-rev? string +--ro firmware-rev? string +--ro software-rev? string +--ro serial-num? string +--ro mfg-name? string +--ro asset-id? string +--ro is-fru? boolean +--ro mfg-date? yang:date-and-time +--ro uri* inet:uri¶
For state data like admin-state, oper-state and so on, we consider they are related to device hardware management and not hardware inventory. Therefore, they are outside of scope of this document. Same for the sensor-data, they should be defined in some other performance monitoring data models instead of inventory data model.¶
We re-defined some attributes listed in [RFC8348], based on some integration experience for network wide inventory data.¶
[RFC8348] provided a "parent-ref" attribute, which was an identifier reference to its parent component. When the MDSC or OSS systems want to find this component's grandparent or higher level component in the hierarchy, they need to retrieve this parent-ref step by step. To reduce this iterative work, we decided to provide a list of hierarchical parent components' identifier references.¶
+--ro components +--ro component* [uuid] ................................... +--ro parent-references | +--ro equipment-room-uuid? leafref | +--ro ne-uuid? leafref | +--ro rack-uuid? leafref | +--ro component-references | +--ro component-reference* [index] | +--ro index uint8 | +--ro class? leafref | +--ro uuid? leafref ...................................¶
The hierarchical components' identifier could be found by the "component-reference" list. The "index" attribute is used to order the list by the hierarchical relationship from topmost component (with the "index" set to 0) to bottom component.¶
According to the management requirements from operators, some important attributes are not defined in [RFC8348]. These attributes could be component-specific and are not suitable to define under the component list node. So, we defined a choice-case structure for this component-specific extension, as follows:¶
+--ro components +--ro component* [uuid] ................................... +--ro (component-class)? +--:(chassis) | +--ro chassis-specific-info +--:(container) | +--ro slot-specific-info +--:(module) | +--ro board-specific-info +--:(port) +--ro port-specific-info ...................................¶
Note: The detail of each *-specific-info YANG container is still under discussion, and the leaf attributes will be defined in future.¶
According to the description in [RFC8348], the attribute named "model-name" under the component, is preferred to have a customer-visible part number value. "Model-name" is not straightforward to understand and we suggest to rename it as "part-number" directly.¶
+--ro components +--ro component* [uuid] ................................... +--ro part-number? string ...................................¶
Note: add some more attributes about equipment room in the future.¶
Besides the common attributes mentioned in above section, rack could have some specific attributes, such as appearance-related attributes and electricity-related attributes. The height, depth and width are described by the figure below (please consider that the door of the rack is facing the user):¶
The rack attributes include:¶
+--ro racks +--ro rack* [uuid] ................................... +--ro height? uint16 +--ro width? uint16 +--ro depth? uint16 +--ro max-voltage? uint16 ...................................¶
Max-voltage: the maximum voltage supported by the rack.¶
We consider that some of the attributes defined in [RFC8348] for components are also applicable for network element. These attributes include:¶
+--ro network-elements +--ro network-element* [uuid] ................................... +--ro hardware-rev? string +--ro firmware-rev? string +--ro software-rev? string +--ro mfg-name? string +--ro mfg-date? yang:date-and-time +--ro part-number? string +--ro serial-number? string +--ro product-name? string ...................................¶
Note: Not all the attributes defined in [RFC8348] are applicable for network element. And there could also be some missing attributes which are not recognized by [RFC8348]. More extensions could be introduced in later revisions after the missing attributes are fully discussed.¶
During the integration with OSS in some operators, some efficiency/scalability concerns have been discovered when synchronizing network inventory data for big networks. More discussions are needed to address these concerns.¶
Considering that relational databases are widely used by traditional OSS systems and also by some network controllers, the inventory objects are most likely to be saved in different tables. With the model defined in current draft, when doing a full synchronization, network controller needs to convert all inventory objects of each NE into component objects and combine them together into a single list, and then construct a response and send to OSS or MDSC. The OSS or MDSC needs to classify the component list and divide them into different groups, in order to save them in different tables. The combining-regrouping steps are impacting the network controller & OSS/MDSC processing, which may result in efficiency/scalability limitations in large scale networks.¶
An alternative YANG model structure, which defines the inventory objects directly, instead of defining generic components, has also been analyzed. However, also with this model, there still could be some scalability limitations when synchronizing full inventory resources in large scale of networks. This scalability limitation is caused by the limited transmission capabilities of HTTP protocol. We think that this scalability limitation should be solved at protocol level rather than data model level.¶
The model proposed by this draft is designed to be as generic as possible so to cover future special types of inventory objects that could be used in other technologies, that have not been identified yet. If the inventory objects were to be defined directly with fixed hierarchical relationships in YANG model, this new type of inventory objects needs to be manually defined, which is not a backward compatible change and therefore is not an acceptable approach for implementation. With a generic model, it is only needed to augment a new component class and extend some specific attributes for this new inventory component class, which is more flexible. We consider that this generic data model, enabling a flexible and backward compatible approach for other technologies, represents the main scope of this draft. Solution description to efficiency/scalability limitations mentioned above is considered as out-of-scope.¶
Note: review in future versions of this document whether the component list should be under the network-inventory instead of the network-element container.¶
Note that in [RFC8345], topology and inventory are two subsets of network information. However, considering the complexity of the existing topology models and having a better extension capability, we define a separate root for the inventory model. We will consider some other ways to do some associations between the topology model and inventory model in the future.¶
Note: review in future versions of this document whether network hardware inventory should be defined as an augmentation of the network model defined in [RFC8345] instead of under a new network-inventory root.¶
The proposed YANG data model has been analysed so far to cover the requirements and use cases for Optical Network Inventory.¶
Further analysis of requirements and use cases is needed to extend the applicability of this YANG data model to other types of networks (IP and microwave) and to identify which aspects are generic and which aspects are technology-specific for optical.¶
Figure 3 below shows the tree diagram of the YANG data model defined in module ietf-network-inventory.yang (Section 4).¶
<Add any manageability considerations>¶
<Add any security considerations>¶
<Add any IANA considerations>¶
Since more and more devices can be managed by domain controller through OpenConfig, to avoid that our inventory data model cannot cover these devices' inventory data, we have compared our inventory data model with the openconfig-platform.yang which is the data model used to manage inventory information in OpenConfig.¶
Openconfig-platform data model is NE-level and uses a generic component concept to describe its inner devices and containers, which is similar to ietf-hardware model in [RFC8348]. Since we have also reused the component concept of [RFC8348] in our inventory data model, we can compare the component's attributes between openconfig-platform and our model directly , which is stated below:¶
Attributes in oc-platform | Attributes in our model | remark |
---|---|---|
name | name | |
type | class | |
id | uuid | |
location | location | |
description | description | |
mfg-name | mfg-name | |
mfg-date | mfg-date | |
hardware-version | hardware-rev | |
firmware-version | firmware-rev | |
software-version | software-rev | |
serial-no | serial-num | |
part-no | part-number | |
clei-code | TBD | |
removable | is-fru | |
oper-status | state data | |
empty | contained-child? | If there is no contained child, it is empty. |
parent | parent-references | |
redundant-role | TBD | |
last-switchover-reason | state data | |
last-switchover-time | state data | |
last-reboot-reason | state data | |
last-reboot-time | state data | |
switchover-ready | state data | |
temperature | performance data | |
memory | performance data | |
allocated-power | TBD | |
used-power | TBD | |
pcie | alarm data | |
properties | TBD | |
subcomponents | contained-child | |
chassis | chassis-specific-info | |
port | port-specific-info | |
power-supply | TBD | |
fan | Fan is considered as a specific board. And no need to define as a single component | |
fabric | TBD | |
storage | For Optical and IP technology, no need to manage storage on network element | |
cpu | For Optical and IP technology, no need to manage CPU on network element | |
integrated-circuit | board-specific-info | |
backplane | Backplane is considered as a part of board. And no need to define as a single component | |
software-module | TBD | |
controller-card | Controller card is considered as a specific functional board. And no need to define as a single component |
As it mentioned in Section 2.1.2 that state data and performance data are out of scope of our data model, it is same for alarm data and it should be defined in some other alarm data models separately. And for some component specific structures in openconfig-platform, we consider some of them can be contained by our existing structure, such as fan, backplane, and controller-card. And for some of them, there is no need to manage for operators, such as storage and cpu.¶
Mostly, our inventory data model can cover the attributes from OpenConfig.¶
The authors of this document would like to thank the authors of [I-D.ietf-teas-actn-poi-applicability] for having identified the gap and requirements to trigger this work.¶
This document was prepared using kramdown.¶