ATSSS the Future of Wi-Fi and Cellular Convergence

Will new and better technology and standards for automatic network selection and intelligent convergence between mobile and Wi-Fi services be developed for the mass market of the future? The short answer is probably yes. We will address one of them here, namely the newly released Access Traffic Steering, Switching & Splitting (ATSSS) introduced in 3GPP release 16.

But the answer is also that such technologies – including Passpoint with SIM authentication – already exist for the most part. These may not be ideal but are still extensively field-proven and work well enough to have already been implemented by dozens of major carriers.

Operators actively choosing Wi-Fi offload as a strategy and who want more granular control often include so-called connectivity manager clients (apps or hidden clients) on the device. Such solutions can be pretty sophisticated depending on to what extent the app, and hence the operator, can access and control the communication layer in the device’s operating system.

The capability of such apps or hidden clients must at least include solutions to the following current imperfections in switching between Wi-Fi and mobile network access:

• Avoiding unintentional ‘walk-by’ switchover to public Wi-Fi, which could produce a poor user experience or even intermittent loss of connectivity.
• Policies and thresholds should automatically reject or accept handoff to Wi-Fi and back to cell sites if either is congested.

The Three “S” in ATSSS

Wouldn’t it be a significant step up in performance and quality of experience if a phone natively could aggregate the data streams from Wi-Fi and cellular into one stream and perhaps even intelligently steer and switch traffic between the two?

We think yes – and fortunately, the 3GPP seems to think so as well since they have introduced ATSSS as part of the 3GPP Release 16 standard for 5G.

Steering

Choosing the best available network based on speed, cost, and latency.

Switching

Moving seamlessly between 5G and Wi-Fi networks.

Splitting

Splitting the traffic over 5G and Wi-Fi, the split can be set by policies.

ATSSS uses the so-called Multipath TCP (MPTCP) technology, described in our white paper, to allow IP data traffic to flow simultaneously over Wi-Fi and 5G networks. The results are higher data rates, improved overall quality, and even gapless handovers between Wi-Fi and 5G.

Since very few applications and web servers support MPTCP, the ATSSS specifies an MTCP Proxy implemented in the 5G core User Plane Function (UPF). It also defines an ATSSS low layer functionality (ATSSS-LL) to support other protocols such as UDP.

The introduction of ATSSS is excellent news for advanced Wi-Fi service management platforms such as Enea Aptilo SMP, as it makes policy management so much more complex.

ATSSS Steering Modes

These functions, the three “S”, translate to four ATSSS standard steering modes that need to be supported in the device and in the Mobile Core (UPF).

Active-standby

One access network – cellular or Wi-Fi – is the active (default) access network. The traffic is routed over this access network until it becomes unavailable, in which case traffic switches over to the other access network. When the active access network is available again, the traffic is switched back.

Smallest Delay

Traffic is sent over the access network with the smallest delay. The Performance Measurement Function (PMF) determines the latency of each network connection. The underlying multipath protocol can also provide measurements.

Load Balancing

This specifies a fixed percentage for the fraction of the traffic that should connect over the 3GPP network with the rest of the traffic sent on the non-3GPP network. This mode only applies to the quality of service (QoS) flows with a non-guaranteed bit rate (non-GBR).

Priority-based

Traffic is transmitted over a specified high-priority access network (Wi-Fi or cellular). If this access network becomes congested, the traffic overflows onto the other access network. If the high-priority access network becomes unavailable, traffic switches to the other access network (as in Active Standby). The determination of congestion is implementation-specific.

Another factor that adds to the complexity of policy management is the large number of stakeholders. Real-world deployment of ATSSS will need to cater to:

• Service provider policies
• Policies set by the user
• Device vendor policies
• App provider policies
• Enterprise IT policies

We think that ATSSS is a very promising standard. It is, to some extent, the ‘Holy Grail’ of mobile data offload, and with ATSSS, operators may finally find a good reason for backhauling Wi-Fi traffic to the mobile core. However, no 3GPP standard for Wi-Fi integration will ever be implemented in practice unless the device vendors want it.

No Reason to Wait for ATSSS

For ATSSS to reach the mass market, device support is crucial. An example of a related standard that never achieved any market penetration at all is 3GPP ANDSF, which was a useful concept but, in the end, was never implemented natively in any device.

It may take quite a few years more for ATSSS to come to market – or alternatively, proprietary forms of essentially the same function incorporated by Apple or others may, in the end, supersede the 3GPP’s attempts. The ATSSS concept has already been tested successfully by Korea Telecom using a proprietary solution.

In either case, there is a good likelihood that Wi-Fi and 5G data streams will find new ways of complementing each other – including using aggregation & gapless handovers – on the transport layer.

Meanwhile, all the benefits of known and field-proven systems for cellular and Wi-Fi convergent services remain available to any operator who wishes to apply vastly improved Wi-Fi technology as a part of their network strategy today. Passpoint and EAP-SIM-based solutions are readily available and can possibly be complemented with an app for more granular control. In other words: Even though a more systematic 3GPP-based approach to convergence may emerge in the coming years, there is no reason to wait. Excellent convergence solutions exist today.