By Thierry Van de Velde, Consulting Technology Specialist @ Alcatel-Lucent
On September 17 there was quite some excitement here at Alcatel-Lucent as we filed for patent for native bridged communication in a cellular access network (EP15306442). It is now entering an examination phase of course, and will need to be defended in the coming months.
This breakthrough is the result of months of R&D by our Wireless colleagues, Bell Labs and the IPRT division.
Our goal is to apply it in 5G and augmented 4G+ networks, but what does it mean to the Carrier Wi-Fi industry and to the Wi-Fi industry in general?
Wi-Fi, Ethernet, fixed broadband wholesale, microwave links, datacenter fabrics, SDN, SD-VPN, E-Line, E-LAN, Ethernet VPN… nearly all modern communication networks offer layer 2 (bridged or switched) access to their users.
Since cellular networks only offer layer 3 (IP) access, today to transport a layer 2 frame across a 2G/3G/4G network, the UE would need to encapsulate (wrap) the frame into an IP packet. This would bring numerous drawbacks : lower throughput, higher battery consumption, needless overhead, packet fragmentation and last but not least the fact that the cellular network can no longer identify nor differentiate the applications, for example voice or video media streams requiring to be prioritized over data traffic.
Moreover without being able to inspect the traffic inside the tunnel, a hybrid network can no longer spread traffic among 3GPP access and non-3GPP access (while avoiding out-of-sequence delivery). Tunnelled traffic to a UE is then conveyed via either the 3GPP or the non-3GPP access connection, but not both.
So why aren’t cellular networks supporting a native Layer 2 service today? Why hasn’t this happened in 2G, 3G and 4G for the last 2 decades? The root cause lies in circuit-switched telephony. 2G networks had been designed for Circuit-Switched services using the International Mobile Subscriber Identifier (E.212 IMSI) as the private identity and the Mobile Subscriber Integrated Services Digital Network Number (MSISDN) as the public identity.
Data services such as SMS and USSD had also been built on these identities. When GPRS was introduced in 3GPP Release 97 nobody saw the need to change the User Equipment (UE) identifiers, since it was a General Packet Radio Service and not a General Framed Radio Service. Nobody saw the need to produce, configure or assign a Layer 2 address to the UE.
Earlier this year Bell Labs uncovered the possibility for the UE to derive a fairly stable L2 address from the Globally Unique Temporary Identity (GUTI) hence indirectly from the IMSI, rather than having to burn a permanent L2 address into the UE’s cellular chipset. That was a masterpiece and is subject to a separate patent application. We further expanded the algorithm to devices not containing a SIM but a security certificate – the identifier already used for Wi-Fi Hotspot 2.0.
Last month 3GPP SA1 took a historic decision, namely to open up 5G networks for access by devices no longer containing a SIM. The battle is not over and the SIM card lobby may kick in but a historic taboo just took a hit.
Besides native bridged communication and SIM-less access a third idea is rapidly gaining ground in the cellular industry, namely that 5G networks (and augmented 4G+) should offer a contention-based, connectionless user plane channel on the air interface, just like Wi-Fi and Ethernet – their main competitors. Today’s LTE networks only offer user plane channels which are dedicated per UE and per QoS Class (QCI).
So will 5G networks become Wi-Fi-like? Not at all. The 5G air interface will continue to offer common control channels, dedicated control channels and dedicated user plane channels as in 4G, besides the new contention-based channel in the user plane. The new 5G waveform (UF-OFDM) will greatly improve spectral efficiency over Wi-Fi and 4G, especially at cell edge. An average of 5 bits/s/Hz per cell is not an unrealistic target. 5G RAN will offer and enforce a rich palette of QoS options not just in the user plane but also in the control plane.
5G core networks will be able to partition their resources among dedicated and shared channels, among different groups of users, with various degrees of mobility. They will permit inter-UE communication like a LAN, VLAN or WLAN but in a controlled fashion, for example only for devices belonging to a single family or company. 5G networks will be able to page the user and to adapt the paging strategy to the use case – static sensors vs. mobile broadband devices. 5G networks will offer numerous structural advantages over Wi-Fi and Ethernet, even at low power (100mW) in unlicensed frequency bands.
In 4G networks it takes an eternity (50-100ms) for a UE to wake up from low-power state. During a tedious procedure with over 20 signalling messages a new security association must be set up between the UE and the (formerly used or new) eNodeB, for encryption on the air interface. Did you know that your 4G smartphone is performing this procedure every couple of seconds, for every burst of packets it wants to send? How could we ever reduce round-trip delays if we keep setting up such dedicated user plane channels and cryptotunnels between UE and RAN?
The hybrid 5G/4G+/Wi-Fi network will be able to encrypt traffic to the UE in a core network node, not only permitting signallingless wake-up, but also very efficient data-triggered handovers at spectacular speeds, and smart distribution of traffic between all available RAN nodes or access technologies offering quality coverage to the UE (3GPP and non-3GPP).
In that Next-Generation Mobile Core (NGMC) we are also taking the opportunity to separate a control plane node from a user plane node communicating via a standard protocol. The user plane node should concentrate on hybrid access, per-UE encryption and low round-trip delays to the content, contrary to today’s SGW, PGW, TWAG or BRAS/BNG. We will ensure independent scaling of central control plane nodes (Hybrid Access Controllers) and distributed user plane nodes (Hybrid Access Gateways).
In actual implementations the HAC will most likely be virtualized whereas the HAG are either integrated in distribution, aggregation or core routers, or in edge computing nodes running both vRAN and vHAG.
The NGMC would grant access to SIM-based and non-SIM-based devices using the Extensible Authentication Protocol (EAP) just like today’s Wi-Fi Hotspots 2.0. Therefore unlike the MME the HAC needs access to a Diameter Authentication, Authorization & Accounting (AAA) server and to other northbound policy control systems permitting fair partitioning of 5G resources – user plane but also control plane.
Access to the Evolved Packet Core (EPC) via S1 or S2a is only possible for SIM-based devices and is only required when more advanced, routed or NATed services are offered.
Augmented 4G (4G+), non-3GPP access networks (Trusted Wi-Fi, Fixed Broadband) and datacenters (SDN platforms, Virtual Machines and Containers) would be connected to the same NGMC, as illustrated on the following diagram :
A single UE may of course be connected to both 5G and Wi-Fi or 4G+ and non-Wi-Fi. The 5G/4G+ RAN, HAC and HAG will have a real-time view on the available bandwidth and quality metrics over the different access technologies, and adapt traffic distribution accordingly for concurrent access.
The future of Carrier Wi-Fi is to become a Trusted Wireless Access technology to a hybrid, multi-technology Next-Generation Mobile Core (NGMC) offering native bridged communication.
Native bridged communication will enable fantastic new services such as: