| Internet-Draft | Applicability Statement | October 2024 |
| Zhang, et al. | Expires 24 April 2025 | [Page] |
Time-Variant Routing (TVR) is a routing system that can support the predicted topology changes caused by internal or external reasons. Typical use cases include mobility networks with moving nodes, resource constraints such as power, and tidal traffic demand networks. This document provides examples of how to implement the TVR scheduling capabilities for key use cases. It describes which part of the TVR data model is used and why, and it outlines operational and security considerations when deploying TVR-based technologies.¶
This note is to be removed before publishing as an RFC.¶
Discussion of this document takes place on the Time-Variant Routing Working Group mailing list (tvr@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/tvr/.¶
Source for this draft and an issue tracker can be found at https://github.com/zhangli-abcd/TVR-Applicability-2.¶
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 24 April 2025.¶
Copyright (c) 2024 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.¶
The Time-Variant Routing (TVR) Working Group addresses a growing need in modern network environments where predictable variations in topology - such as the restoration, activation, or loss of network elements, are part of normal operations. This approach is essential in dynamic networks with mobile nodes, where links may be frequently disrupted and re-established due to mobility or in networks with highly predictable traffic patterns, where links may be powered down to conserve or reduce energy.¶
This document provides examples of implementing TVR scheduling capabilities in identified use cases. It demonstrates the applicability of the TVR data model, methods for disseminating the TVR schedule, and the necessary IETF ancillary technologies for network environments, such as time synchronization and policy, that support TVR capabilities.¶
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 tidal network is a typical scenario of an Energy Efficient case (Section 3 of [I-D.ietf-tvr-use-cases]). The tidal network means that the volume of traffic in the network changes periodically, like the ocean tide. These changes are mainly affected by human activities. Therefore, this tidal effect is obvious in human-populated areas, such as campuses and airports.¶
In the context of a tidal network, if the network maintains all the devices up to guarantee a maximum throughput all the time, often, power will be wasted on network resources that are not being used. Energy-saving methods may include the deactivation of some or all components of network nodes. These activities have the potential to alter network topology and impact data routing/forwarding in a variety of ways. Interfaces on network nodes can be selectively disabled or enabled based on traffic patterns, thereby reducing the energy consumption of nodes during periods of low network traffic. The following sections will describe how to use the TVR YANG model[I-D.ietf-tvr-schedule-yang] in Tidal Network.¶
According to the description of Section 3.1 of [I-D.ietf-tvr-requirements], the scheduling generation locality and execution locality may be centralized or distributed.¶
When the schedule is generated and executed in a centralized manner, the changes are sent to routing applications in wall-clock time via a management interface, which does not need to be delivered by the YANG model. This can be implemented using the existing management plane technology. Therefore, this scenario is outside of the scope of this document.¶
When the schedule is generated and executed in a distributed manner, there is also no need for delivering the schedule by the YANG model. Therefore, this scenario is also outside of the scope of this document.¶
When a schedule is generated in a centralized manner and executed in a distributed manner, the YANG module needs to be used to deliver the schedule to the managed device. Depending on the location of the routing application, the network model can be divided into two types: distributed and centralized.¶
In the centralized model, the network managing device generates and maintains schedules, the routing application is deployed in the network controller, and the network devices execute the schedules and routing results. The following figure shows the components of the centralized model.¶
+-----------------------------------------------------------------+
| Managing Device |
+-----------------------------------------------------------------+
| |
Data Model Data Model
| |
+---------\|/----------+ +--------\|/--------+
| Network Controller |---Routing Results-->| Network Devices |
+----------------------+ +-------------------+
In the distributed model, the managing device generates and maintains schedules, the routing application is deployed in the network devices which also executes schedules and route calculation.¶
+-------------------------------------------------+
| Managing Device |
+-------------------------------------------------+
|
Data Model
|
+-----------------------\|/-----------------------+
| Managed Device (Network Devices) |
+-------------------------------------------------+
A centralized model involves the interaction between the managing device and network controller, the managing device and network devices, and the controller and network devices.¶
The managing device needs to deliver node-specific schedules to network devices by TVR Schedule Node YANG module Section 5.2 of [I-D.ietf-tvr-schedule-yang], and the network devices need to report their own status data to the management device.¶
The managing device needs to deliver schedules of the network topology to the network controller by the TVR Network Topology YANG moduleSection 5.3 of [I-D.ietf-tvr-schedule-yang], so that the routing application in the controller can consider the impact of topology changes on routes when calculating routes.¶
The network controller should deliver the route calculation result to the network devices. The format of the routing results depends on the protocols deployed(The typical protocols include BGP , PCEP, etc.). The routing result for a period of time in the future could be sent to the network devices in wall-clock time or be packed and sent at some special points.¶
The distributed model only involves the interaction between managing devices and network devices.¶
The managing device must deliver node-specific schedules to network devices by TVR Schedule Node YANG Module and deliver network topology schedules to all the network devices by TVR Network Topology Yang module for route calculation. The network devices must report their status data to the managing device. Editors note: In future versions of this document, we will provide a more detailed procedure for both the distributed and centralized approach.¶
The TVR data model [I-D.ietf-tvr-schedule-yang] can manage network resources and topologies with scheduled attributes. There are modules defined in the TVR data model, these are:¶
The “ietf-tvr-schedule” module contains the schedule YANG definitions. This module uses groupings from [I-D.ietf-netmod-schedule-yang] data model;¶
The “ietf-tvr-topology” module defines a network topology with a time-variant availability schedule;¶
The “ietf-tvr-node” module is to be used to manage the scheduled attributes of a single node.¶
To create a schedule, the following TVR data model objects and subsequent branches are used:¶
A TVR scenario example is provided below, where a wireless link is shut down for 12 hours, from 19:00 to 7am the next day. The schedule is identified using a unique identifier that is conveyed in ‘schedule-id’, and the recurring schedule can be applied for multiple days using Coordinated Universal Time (UTC). A more detailed example of the json code is provided in this documents Appendix.¶
{
"ietf-tvr-node:node-schedule":[
{
"node-id":1234567890,
"node-power-schedule":{
"power-default":true,
},
"interface-schedule":[
{
"name":"Wlan0",
"default-available":false,
"attribute-schedule":{
"schedules":[
{
"schedule-id":111111,
"recurrence-first":{
"utc-start-time":"2025-12-01T19:00:00Z",
"duration":43200
},
"utc-until":"2026-12-01T00:00:00Z",
"frequency":"ietf-schedule:daily",
"interval":1,
"attr-value":{
"available":true
}
}
]
}
}
]
}
]
}
¶
The methods for disseminating and propagating the generated schedule are discussed in the following subsections.¶
The TVR data model is designed to be accessed via YANG-based management protocols such as NETCONF [RFC6241] and RESTCONF [RFC8040]. This section discusses the applicability of these protocols for configuring time-variant network resources using the TVR YANG data models.¶
NETCONF provides a robust mechanism for managing complex network configurations, particularly when transactional integrity and operational consistency are required. Its ability to execute atomic transactions ensures that schedules involving multiple resources are applied fully, preventing partial updates that could lead to configuration inconsistencies. This feature is important for time-sensitive scheduling in TVR environments. Additionally, NETCONF supports the validation of configurations prior to commitment, allowing operators to verify the correctness of schedules before they are applied. It also includes rollback capabilities, such as restoring a previous configuration during scheduling errors.¶
In contrast, RESTCONF offers a simpler, stateless method for interacting with network devices, making it suitable for use cases requiring lightweight, rapid configuration. RESTCONF utilizes a RESTful interface over HTTP, providing a streamlined approach to network configuration and management. Therefore, RESTCONF may be advantageous in scenarios where quick adjustments to schedules are needed or where integration with web-based or cloud-native systems is a priority.¶
Depending on the type of node in the TVR network, NETCONF would be the preferred protocol for large-scale, critical scheduling operations requiring validation and rollback mechanisms. For smaller-scale or isolated scheduling tasks, RESTCONF provides an efficient and straightforward option without the need for the transactional features offered by NETCONF. The choice of protocol to use with the TVR YANG model should be driven by the specific requirements of the network environment and the complexity of the scheduling tasks involved.¶
The security aspects of both NETCONF and RESTCONF, including their strengths and weaknesses, are discussed further in Section 8 of this document.¶
According to Section 3.1.3 of [I-D.ietf-tvr-requirements], no matter whether the schedules are executed in a centralized or distributed mode, a mechanism is required to keep the time synchronization between different devices.¶
Different time-variant scenarios may require different granularities of time synchronization. For example, the period of traffic and topology changes in tidal networks is usually a day or week. Therefore, a second-level time synchronization is enough.¶
Existing clock synchronization protocols can be classified into hardware-based protocols and software-based protocols. Hardware-based protocols often rely on dedicated hardware to ensure clock synchronization, such as Global Positioning System (GPS) and Precision Time Protocol (PTP). Software-based protocols, on the other hand, synchronize clocks through software packages running on systems, such as Network Time Protocol (NTP) [RFC5905] and Simple Network Time Protocol (SNTP) [RFC4330].¶
Hardware-based protocols typically have higher precision and stability, but also have higher cost due to the dedicated Hardware. Software-based protocols are simple and applicable to common hardware devices, but have lower precision (For example, the NTP can realize the synchronization at tens of milliseconds level). The software-based protocols are appropriate for most of the TVR use cases.¶
NTP is fundamental for ensuring that TVR mechanisms, which depend on highly accurate timing, function as intended across an entire network. Misalignment in time could lead to serious routing issues, including inefficiency in path forwarding, instability in routing tables, and traffic outages.¶
NTP will be used to ensure:¶
Coordination of Planned Network Events;¶
Verification of TVR Data Model Time Stamps¶
Accurate Scheduling of Paths;¶
Fault Tolerance.¶
NTP uses a hierarchical structure of time sources. Each level of this hierarchy is termed a stratum. Generally, an NTP server synchronized to an authoritative clock runs at stratum 1. Such NTP server functions as the primary time server to provide clock synchronization for other devices on the network. Stratum 2 servers obtain time from stratum 1 servers, stratum 3 servers obtain time from stratum 2 servers, and so on.¶
In TVR use cases, the managing device functions as a level-1 NTP server and synchronized to an authoritative clock source. The network controller and network devices behave as clients to obtain accurate time from the managing device. Figure 3 shows an NTP deployment scenario for obtaining clock from a GPS clock source.¶
+--------------------+
| GPS Clock Source |
+--------------------+
|
+--------------------\|/------------------+
Stratum 1 | Managing Device |
+-----------------------------------------+
| |
| |
| |
+---------\|/----------+ +--------\|/--------+
Stratum 2 | Network Controller | | Network Devices |
+----------------------+ +-------------------+
SNTP is a subset of the NTP used to synchronize computer clocks in the Internet. It simplifies the complex NTP synchronization function and is suitable for networks with limited resources and loose precision requirements. Compared with NTP, SNTP has lower clock precision, but the synchronization precision still can be guarded under seconds. Therefore, SNTP also meets the time synchronization requirements of tidal networks and can be used as an alternative clock synchronization protocol.¶
The schedule database is used to store and maintain schedules, the database may be deployed on a managing device and managed devices based on requirements.¶
The schedule source of the schedule database may be diversified, for example, configuration from an administrator or YANG model from the management interface. The schedule entries of different databases may be different, but the content of the same schedule entry in the schedule databases of different devices in the same domain must be consistent. There are at least two ways to make the content of the same schedule entry in different schedule databases consistent:¶
All the schedule entries are generated at a specific device;¶
Schedule entries are generated at different devices, but there is a synchronization mechanism to synchronize the schedule databases among devices.¶
Option 1 is simplest and easy to implement. In a time-variant domain, the managing device may receive scheduling requests and generate all schedule entries. Then the schedule entries are delivered to the necessary network devices in the domain through the TVR YANG model.¶
Option 2 relies on advertisement mechanisms (such as routing techniques) to advertise the scheduling data generated by itself to other devices. This could be achieved using extensions to existing routing schemes and techniques.¶
These options will be discussed with the TVR Working Group, and agreed approaches will be documented in future versions of this Internet Draft.¶
[I-D.ietf-tvr-schedule-yang] defines a TVR Node and TVR Topology YANG modules. The Node YANG module includes node power schedule and interface schedule. The Topology YANG module includes nodes schedule and links schedule.¶
Based on the preceding four schedule types, the schedule database should contain four types of schedule entries in different formats:¶
The detailed format and fields of different types of schedule entries could reference to the definitions of the corresponding YANG modules.¶
Editors note: Code examples will be provided here in future versions of this document.¶
This section provides general requirements for using the TVR schedule.¶
The schedule database should support the add, update, and delete operations.¶
When adding or updating a schedule entry, the execution node needs to check whether resource conflicts exist between the current schedule and existing schedules. If a conflict exists, the operation should be failed.¶
Schedules are updated and deleted based on schedule IDs. Therefore, schedule IDs must be unique in a time-variant domain. This can be handled, e.g., by a dedicated allocation agent within the time-variant domain.¶
Editors note: Future versions of this document will expand on the schedule operations requirements and best practices.¶
Several operational considerations exist when using TVR techniques and data models in a network. This section provides some high-level observations and more detailed sub-sections for specific consideration related to schedule dissemination, execution, and recovery in case of failure to apply a schedule or partial change.¶
Coordinated Network Events: TVR often coordinates routing changes anticipating events like predictable low-traffic periods or link downtimes (e.g., scheduled maintenance or traffic demand).¶
Accurate Scheduling of Paths: TVR schedule capable routers and network nodes will dynamically adjust forwarding paths based on planned changes in link availability or network conditions.¶
Time-Stamped Data Models: TVR will require the use time-stamped data models (e.g., schedules for link changes or availability windows) to make interface management decisions. This ensures that all TVR nodes interpret the timing of events consistently and implement time-based policies correctly.¶
Therefore, network time accuracy and time-stamped data models are critical to ensure that coordinated network events and scheduled path decisions across the network are based on a consistent time reference. Without accurate time sync, nodes could apply different schedules, causing routing inconsistencies, path flapping, or packet loss.¶
Schedules execution means that a component (e.g., device) undertakes an action (e.g., allocates and deallocates resources) at specified time points. In a tidal network, the schedule execution indicates powering on/off specific network components (such as interfaces or entire network devices) directly or by commands.¶
The schedule executor should understand the consequences of the schedule execution. The power on/off of network components usually affects the network topology, the addition and deletion of the topology need to be considered separately.¶
A link coming up or a node joining a topology should not have any functional change until the change is proven to be fully operational. The routing paths may be pre-computed but should not be installed before all of the topology changes are confirmed to be operational. The benefits of this pre-computation appear to be very small. The network may choose to not do any pre-installation or pre-computation in reaction to topological additions at a small cost of some operational efficiency.¶
Topological deletions are an entirely different matter. If a link or node is to be removed from the topology, then the network should act before the anticipated change to route traffic around the expected topological change. Specifically, at some point before the planned topology change, the routing paths should be pre-computed and installed before the topology change takes place. The required time to perform such planned action will vary depending on the exact network and configuration. When using an IGP or other distributed routing protocols, the affected links may be set to a high metric to direct traffic to alternate paths. This type of change does require some time to propagate through the network, so the metric change should be initiated far enough in advance that the network converges before the actual topological change.¶
The integration of time-variant mechanisms in network operations presents distinct security challenges that require thorough analysis to safeguard the network’s integrity, availability, and confidentiality. Networks that rely on time-sensitive data for routing and forwarding decisions are particularly susceptible to attacks that exploit timing dependencies.¶
The "Security Considerations" section of [I-D.ietf-tvr-requirements] outlines various threat vectors and categories specific to time-variant environments. In this document, a tidal demand use case has been presented, highlighting the need for robust security measures in the processes of generating, disseminating, and applying schedules to control tidal interfaces. It is essential to assess security risks and implement mitigation strategies to ensure the security of both the network and individual nodes throughout the scheduling process.¶
Future iterations of this document will provide detailed security techniques and best practices to address these challenges.¶
This document has no IANA actions.¶
TODO acknowledge.¶
Tidal networks usually need to manage the availability of some node or interfaces. Figure 4 shows the example of a scheduling node that is powered on from 12 AM, December 1, 2025 to 12 AM, December 1, 2026 in UTC and its interface named "interface1" is scheduled to be enabled at 7:00 AM and disabled at 1:00 AM, every day, from December 1, 2025 to December 1, 2026 in UTC.¶
The JSON encoding is used only for illustration purposes.¶
{
"ietf-tvr-node:node-schedule":[
{
"node-id":12345678,
"node-power-schedule":{
"power-default":false,
"schedules":[
{
"schedule-id":111111,
"period-start":"2025-12-01T00:00:00Z",
"period-end":"2026-12-01T00:00:00Z",
"attr-value":{
"power-state":true
}
}
]
},
"interface-schedule":[
{
"name":"interace1",
"default-available":false,
"default-bandwidth":1000000000,
"attribute-schedule":{
"schedules":[
{
"schedule-id":222222,
"recurrence-first":{
"utc-start-time":"2025-12-01T07:00:00Z",
"duration":64800
},
"utc-until":"2026-12-01T00:00:00Z",
"frequency":"ietf-schedule:daily",
"interval":1,
"attr-value":{
"available":true
}
}
]
}
}
]
}
]
}