Requirements of
applying path-aware networking for dynamic path selection
China Mobile
Beijing
100053
China
fuyuexia@chinamobile.com
China Mobile
Beijing
100053
China
liupengyjy@chinamobile.com
PANRG
Emerging new services have new business characteristics, different
from traditional C/S business model, whose most traffic is downstream
traffic, more and more new business with gradually increasing upstream
traffic have appeared, such as short videos, live sales etc, . Due to
the new traffic characteristics of these services, more requirements
have been put forward for the choice of network paths. In addition,
emerging services also put forward new requirements for computing. Only
selecting the network path or the service node cannot meet the stringent
requirements. The perception of network paths and path selection also
need to consider the characteristics of the service, and further need to
coordinate the state of the network side and the service node side. The
application of path-aware networking can assist the terminal to better
perceive the network status, and also combine the status of the service
node to achieve on-demand, more fine-grained path selection.
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this
document are to be interpreted as described in RFC 2119.
In path-aware networking architecture, endpoints have the ability to
select or influence the path through the network used by any given
packet or flow. The network and transport layers explicitly expose
information about the path or paths available from one endpoint to
another, and to those endpoints and the applications running on them, so
that they can make this selection [draft-irtf-panrg-questions-09]. This
draft targets at the third question in [draft-irtf-panrg-questions-09]:
how can endpoints select paths to use for traffic in a way that can be
trusted by the network, the endpoints, and the applications using
them?
And this draft targets at the path selection use case of path-aware
networking, and we both consider the scenario that a set of paths to the
same destination and also the scenario that several destinations with
several paths. According to [draft-irtf-panrg-path-properties-02],
entities may select their paths to fulfill a specific goal, e.g.,
related to security or performance, as an example of performance related
path selection, an entity may prefer paths with performance properties
that best match its traffic requirements. In this draft, we target at
the services with various traffic requirements for upstream and
downstream traffic and also with requirements to service endpoint
Different types of services have different requirements to
network:
1. For transmission-intensive services, the amount of data
transmitted is large, so the choice of network path affects the entire
service larger.
2. For computing-intensive services, the computing tasks of service
endpoint are complex and the choice of endpoint affect the entire
service is large
3. And traditional transmission-intensive services tend to have a lot
of downstream traffic, so they usually specify the downstream path.
4. For transmission-intensive services with large upstream traffic,
such as short video and live broadcast, the upstream path matters a lot
so the perception and specification of upstream path is necessary to
meet service requirements.
So the terminal needs to be aware of both the status of the uplink
path and the downlink path, and specify the uplink path and the downlink
path based on service characteristics. What’s more, for
computing-intensive services, the terminal still needs to be aware of
the status of service endpoint, and the path-aware networking also need
to consider the status of endpoint when select network path.
For services with different requirements, when path-aware networking
is applied to realize path perception, it is necessary to dynamically
determine the perceived target paths and target path attributes, such as
perceiving the given upstream path or the given downstream path, and
perceiving path latency or path bandwidth
[draft-irtf-panrg-path-properties-02]. When user initiates a service
request, path-aware networking needs to analyses service requirements
related to path-awareness, including time sensitivity, traffic amount,
and traffic characteristics etc, and decide to be aware of which set of
paths and which path properties of them. So path-aware networking needs
to specify the following information:
1. Service requirements towards path-awareness
2. Target paths to be perceived
3. Target path properties to be perceived
For example, when a service with large amount of uplink traffic and
strict requirements on service latency is requested, path-aware
networking assign a set of uplink paths which are to be perceived, and
determine the target path property is path latency, and then specify the
above mentioned upstream paths to user, and then user initiate uplink
path detection packet towards given paths carrying target path
properties , and then the network nodes along the path writes the
required path properties information. With path-aware networking, the
given paths and corresponding properties are obtained, and user can
select optimal uplink path which meet service requirements.
In path-aware network, instead of using single MED value, other
properties such as Link Capacity or Link Usage could additionally be
used to improve load balancing or performance
[I-D.ietf-idr-performance-routing]. And more properties are required to
be considered for new emerging services
[draft-irtf-panrg-path-properties-02].
The transmission of upstream traffic and downstream traffic, and also
data processing by the service endpoint form a complete service process
(face recognition, CLOUD A/VR, etc.). So the completion of the service
needs to consider multi-dimensional factors.
For path-aware networking, facing diverse service requirements and
multi-dimensional path properties, to solve the problem of how to
comprehensive select path considering service requirements, a new
parameter needs to be introduced: path property weight values, which
represent the weight of each path properties and are used to
comprehensively define the perceived multi-dimensional path properties.
And then the path-aware networking needs to specify the following
information:
1. Service requirements towards path-awareness
2. Target paths to be perceived
3. Target path properties to be perceived
4. Path property weight values of target path properties
For example, for requested services that require large uplink
bandwidth, path-aware networking need to define larger uplink path
bandwidth weight, and calculates the target "uplink path + downlink
path" pair based on the given weight value.
Many emerging services not only put forward new requirements for the
network, but also put forward requirements for computing. For services
such as AR/VR, the budgets for computing delay and network delay are
almost equivalent [draft-liu-dyncast-ps-usecases-01] , therefore, when
path-aware network perceives paths, designates path and selects paths,
it also needs to consider the status of the service endpoint. And then
the path-aware networking needs to specify the following
information:
1. Service requirements towards path-awareness, including service
endpoint
2. Target service endpoints and properties
3. Target paths to be perceived corresponding to target service
endpoints
4. Target path properties to be perceived
5. Path property weight values of target path properties including
service endpoint
And when the requested service is a computationally intensive
service, the status of the service endpoint will have a greater impact
in the entire process. Therefore, it is also necessary to select an
optimal service endpoint to provide services. Path-aware networking
needs to generate multiple target paths for multiple candidate service
endpoints, and specify new path parameter weight values towards target
path properties and target service endpoint status.
The dynamic path selection considering service requirements and
service characteristics has become one of the current technical
development directions. This draft analyzes the application of
path-aware networking to achieve the on-demand path awareness and
service endpoint awareness, and provides optimal path selection.
This document makes no request of IANA.
Note to RFC Editor: this section may be removed on publication as an
RFC.