As the mobile industry evolves toward 5G and beyond, it is becoming clear that no single access technology will be able to meet the great variety of requirements for human and machine communications. Multi-access traffic management at the edge is vital for addressing ever increasing performance requirements of today’s and future applications.
Emerging Industry 4.0 applications such as robotics and remote control require extremly low packet loss ratio of less than 10-12 and low latency less than 1ms. However, driving more data through a scarce and finite radio spectrum becomes a real challenge since spectrum efficiency is approaching a plateau, and a single access technology will not be able to deliver the needed increase in data rates.
The following toy example explains the benefits of multi-access edge computing (MEC). Consider N available (wireless) connections between a client and edge, each of which has some data rate (R) and packet loss rate (P). Table 1 shows the linear improvement in data rate, and the exponential improvement in reliability, by seamlessly combining multiple independent connections to deliver user data. However, it is not possible to achieve the maximum data rate (400Mbps) and the highest reliability (10-12) at the same time.
Table 1: Quality of service (QoS) improvement through multi-access (R = 100Mbps, P = 10-3)
|
N = 1 |
N = 2 |
N = 3 |
N = 4 |
---|---|---|---|---|
Data-Rate |
100Mbps |
200Mbps |
300Mbps |
400Mbps |
Reliability |
10-3 |
10-6 |
10-9 |
10-12 |
Hence, the multi-access challenge is to manage data traffic across all available access networks and meet diverse application requirements in rate, latency, and reliability. To address this challenge, the following key questions need to be answered:
1. How to acquire individual application requirements?
2. How to support multi-access traffic management end-to-end?
3. What measurements are needed for making smart decisions?
First of all, applications may have different requirements. For example, a background data-centric application such as email or drop box may not care about delay or throughput, and therefore may prefer a Wi-Fi connection with lower cost. In another example, an ultra reliable and low latency communications (URLLC) application such as robot control and collision avoidance requires very high reliability and low latency, and therefore may prefer sending duplicated packets over multiple connections simultaneously. Also, a real-time high-definition video streaming application requires high throughput, and therefore may prefer splitting traffic over multiple connections through bandwidth aggregation.
Traffic management (TM) service has been recently specified by the European Telecommunication Standards Institute (ETSI) in the ETSI/MEC reference architecture to address differing requirements. It allows applications to get informed of various capabilities and multi-access network connection information, and to provide requirements such as delay, throughput, and loss for influencing traffic management operations at the edge.
Multi-access traffic management must be operated end-to-end, and therefore requires a set of new protocols between client and network. Recently, a multiple access management service (MAMS) has been proposed to address this need. In parallel, the 3rd Generation Partnership Project (3GPP) has developed the access traffic steering, switching, and splitting (ATSSS) feature to enable simultaneous use of one 3GPP and one non-3GPP connection to deliver data flows. Both provide mechanisms for flexible selection of network paths, and leverage network intelligence and policies to dynamically adapt traffic distribution across selected paths under changing network/link conditions.
The multi-access protocol stack consists of two sublayers:
Convergence sublayer: This layer performs multi-access specific tasks such as access (path) selection, multi-link (path) aggregation, splitting/reordering, lossless switching, keep-alive, and probing. A new generic multi-access (GMA) protocol is specified to encode additional control information at this sublayer, including key, sequence number, and time stamp.
Adaptation sublayer: This layer performs functions to handle tunneling, network layer security, and network address translation (NAT). Existing protocols, including user datagram protocol (UDP) and IP security (IPSec), can be reused.
To take full advantage of multi-access connectivity, traffic load should be intelligently distributed across available access links in a manner that improves user experience with efficient radio resource usage. To achieve this goal, measurements that reflect the connectivity conditions of different access networks should be incorporated while determining multi-access traffic distribution.
For example, the end-to-end packet delay measurements can be used to identify which access provides better latency performance. When serving quality of service (QoS) flows requiring high reliability, packet drop ratio measurements give a good indication of whether redundant transmission over multiple access networks is required. In addition to end-to-end packet statistics, radio access network (RAN) measurements can indicate network quality degradation caused by deteriorating radio link quality or access network congestion, in a timely fashion. Example RAN measurements include radio link quality indicators, such as reference signal received power (RSRP), reference signal received quality (RSRQ), and received signal strength indicator (RSSI), and access network utilization levels, such as physical resource blocks (PRB) usage for 3GPP access and basic service set (BSS) load for the wireless local area network (WLAN).
All in all, Intel Labs envisions a multi-access traffic management framework as illustrated in Figure 1, where edge intelligence determines traffic management configurations and informs data plane and client about how to route packets over multiple paths and priority handling for different QoS flows.
Key elements of this framework include:
1. Packet filtering and flow classification: Categorize incoming data traffic into different QoS classes, based on for example ETSI/MEC traffic management interfaces (APIs).
2. Traffic shaping and reliability enhancement: Manage QoS classes with different priority, including rate control and jitter control. In addition, a network coding engine may be applied to high reliability traffic flows to efficiently exploit multipath diversity.
3. Multi-access packet scheduling: Decide which portion of traffic of a multi-connectivity data flow will be routed over each available connection path. This selection is based on end-to-end measurements from the convergence sublayer or/and lower-layer measurements from RAN.
Intel Labs is developing generic multi-access technology based on the multi-access traffic management framework. Using these building blocks, the full potential of multi-access at the edge will be unleashed to address the performance requirements of applications now and in the future. Check out this short demo video clip to get a sneak peek. Stay tuned.
Jing Zhu is a principle engineer with Intel Labs.