Service Mobility-Enabled Computing Aware Traffic Steering using IP address anchoring

Let's consider a possible use case scenario, just for the sake of illustrating the scenario. A terminal is running an AR/VR/XR application (note that this is just an example, other services would also benefit from compute and connectivity traffic steering). Part of this service is executed in the network infrastructure, posing some requirements on the connectivity (e.g., delay between the terminal and the node where the service is executed on the network infrastructure) and computing resources (e.g., capabilities to render the XR video within a certain latency budget). Within the network domain where the terminal is connected to there are multiple sites capable of hosting the service, each with potentially different connectivity and computing characteristics. shows an exemplary scenario. Considering the connectivity and computing latencies (just as an example of metrics), the best service site is #n-1 in the example used in the Figure.

The main problem that this document tries to address is the following. Networking systems do not have mechanisms yet to service mobility mechanisms optimized for connectivity and computing-aware traffic steering, which consider both connectivity and computing status to dynamically select where a service runs for a given terminal. Based on the former, this document proposes solutions to enable the network to react and adapt to connectivity and computing conditions, triggering optimal service migration based on service-specific IP anchor mobility. In particular, this document addresses the following questions: (i) what mechanisms does the network need to implement to facilitate the migration of a service so its requirements in terms of computing and networking are maintained?; and, (ii) how to steer traffic to a new service instance location after moving the service, in a transparent manner to the terminal, by using IP anchor mobility?

We describe next an example of operation and signaling for the network to perform service mobility. Three different embodiments are described next, for variations (OPTIONS) of the procedures: terminal initiated, ECR-initiated and CATS-controller initiated. In addition to the functionality defined in , this documents defines a new functionality:

Service mobility: it deals with the procedures required to (i) detect or predict a change of the current conditions, jointly considering computing and networking, requiring of a service mobility operation; (ii) selecting the best target service instance location, and (iii) triggering the service mobility by orchestrating the service anchor mobility and requesting service migration to a new site. For example, a terminal or ECR might use this functionality to perform active monitoring of a service with CATS agents running at the current ICR, ECR and or service site. It is also used to perform the actual service anchor mobility.

Next, we assume service mobility is triggered by a CATS-aware terminal. By having a CATS agent running on the terminal, it can perform different monitoring actions to predict or detect the need to migrate a service from one site to another. This CATS agent might, for example, interact with other CATS agents deployed on ICRs, ECRs and service sites. In the following we describe a service anchor mobility procedure for CATS, initiated by a CATS-aware terminal. Following the terminal initiation, the network infrastructure is capable to select a target service instance meeting the connectivity and computing requirements of the service, with signaling procedures defined to perform a transparent anchor migration to a new site, facilitating the service migration in a transparent way for the terminal. Extensions and new behavior are highlighted. Note that variations are possible over this exemplary signaling diagram.

| | | | | | | | | | | | | | | | | | | | | | | | |2a.CATS query(service ID, terminal ID, ICR ID, CATS reqs.) | |······>|<··>| | | | | | | | | | |········································>|<··>| | | | 3a.CATS response(service ID, terminal ID, ECR ID, CATS cond.) | |<······| | | | | | | | | | | |<········································| | | | | | | | | | | | | | | | | 4a. Service anchor/ECR@site#n is selected as best | | | | | | | | | | | | | | 5a.CATS request(service ID, terminal ID, ICR ID, CATS reqs., IP pref.) |···············································>| | | | | | | | | | | | | | | | |6.CATS ACK(service ID, terminal ID, ECR ID, CATS cond., IP prefix) |<···············································| | | | | | | | | | | | | | | | | 7.New tunnel for IP pref. established beetween ICR#1 and ECR#N | | | | | | | | | | | | | | | 8.Service migration from site#N-1 to site#N | | | | | | | | | | | | | | | | 9. Service specific traffic| | | | | | | |<---->|<--------------------------------------->|<------>| | | | | | | | | | | | | | ]]> shows the message sequence chart of the IP anchor mobility for CATS, initiated by a CATS-aware terminal (OPTION A), which is explained next:

A service/app has been already instantiated on a service site (for example, by using the mechanisms specified in ). In this example, the site where the service/app consumed by the terminal is currently instantiated is site #n-1. The terminal is connected to ICR #1 and there is a service tunnel between ICR#1 and ECR#N-1. This service/app requires some functionality to be run on the network infrastructure (e.g., an AR/VR/XR service). This service has specific requirements in terms of both connectivity and computing (CATS requirements).
An internal or external trigger is generated regarding a change in the computing of networking conditions, making the current selected service site not feasible for the running service (i.e., CATS requirements cannot be met). In this OPTION A, the terminal is CATS-aware, and it the node that detects the change in the conditions (how this is made is outside the scope of this document, but possible mechanisms include: monitoring at service/app level, in-situ monitoring by the terminal, etc.) and sends a trigger to the ICR.
The ICR sends a query to all (but currently used) ECRs of the domain, or a subset selected based on the location of the ICR. This query may include the following parameters:
1. Service ID: an identifier of the service requested by the terminal. This allows to check if the service can be instantiated or it is already instantiated.
2. Terminal ID: an identifier of the terminal requesting the service. This is useful for example for affinity purposes. It might not include information that can be used to identify the user.
3. ICR ID: identifier of the requesting ICR.
4. CATS requirements: list of requirements, e.g., connectivity and computing requirements.
Each ECR, possibly after checking with the CATS agent of the site(s) it provides connectivity, responds, including the following information:
1. The ICR sends a query to all ECRs of the domain, or a subset selected based on the location of the ICR. This query may include the following parameters:
  1. Service ID.
  2. Terminal ID.
  3. ECR ID: identifier of the ECR sending the response.
  4. CATS conditions: how the site meets each of the requirements included in the request.
  5. (Optional): URI to get to the service instance.
  A CATS agent at a site might be collocated with the ECR. Examples of a CATS agent at a site are network controllers or orchestrators at the site. Note that the way a CATS agent at an ECR may interact with the CATS agent of the site is out of the scope of this document. Examples include using monitoring and telemetry interfaces with an orchestrator managing the site.
2. Based on the received responses, and considering both networking and computing metrics and policies, the ICR selects an ECR (#n).
3. The ICR requests the proposed/selected ECR to establish a traffic steering session with it, sending a CATS request. This request includes the same information that was included in the CATS query (to facilitate stateless operation of the ECRs while being queried), plus the following additional parameters:
  - ECR prefix: currently in use IP prefix IP to the terminal to reach the service instance.
  - Lifetime: requested duration for the association between the ICR and the ECR.
4. The selected ECR responds back with an acknowledgement, including the following information:
  1. Service ID.
  2. Terminal ID.
  3. ECR ID: identifier of the ECR sending the response.
  4. CATS conditions: how the site meets each of the requirements included in the request.
  5. IP prefix assigned for the terminal to use to reach the service instance. It should match the one included in the request.
  6. Lifetime: granted duration of the association between the ICR and the ECR.
5. An IP tunnel is established between the ICR and the new ECR. Optionally (not shown in the figure), the ICR might send a CATS request with cero lifetime to the old ECR to remove the old tunnel.
6. The previous message triggers the service migration (from site #n to site #n-1). The specific mechanism is out of the scope of this document. Note that some preparation/migration steps might be conducted in parallel (e.g., after messages #1 and #3) to accelerate the process, making this step just the final trigger for the service migration. At site #n, the prefix used by the terminal for accessing the service is configured to be used by migrated instance. This might requires routing updates to be perfomed in the site, potentially controlled by a CATS agent running in the site.
7. Traffic of the service for this terminal is steered using the IP tunnel.

TBD.