Releasing the Logjam in the 5G Edge Computing Ecosystem
article by Andy Jones"The future is certain: give us time to work it out" David Byrne, Talking Heads
Predicting the future is a fool’s game. But at least if you’re going to attempt it, you should always try to learn something about the direction of travel based on your experiences so far on the journey. So, let’s begin with a short, potted history from a very personal perspective.
I’ve been consulting and advising up and down the edge computing ecosystem for more than a decade. For edge computing it’s been a long gestation period - but that’s not so unusual for any foundational new technology. Take edge computing’s older sibling - cloud computing, for example. Service-oriented architecture and web services emerged gradually over at least a decade before becoming commercialised in the form of cloud computing offerings such as EC2 from Amazon Web Services. So the precedent exists, however edge is arguably even more challenging an ecosystem to develop than cloud due to the inherent diversity of actors in the value chain.
My journey with what became known subsequently as edge computing began in 2009 while I was still leading network strategy and architecture at Vodafone. My team and I had been allowed some “20% time” and this led to us issuing a Request for Information (RFI) for content caching in the Radio Access Network. The responses to the RFI were not exactly overwhelming. Clearly there were not so many people thinking about edge at that time.
Two years after leaving Vodafone, in 2011 I was consulting for Akamai – the undisputed leader in content delivery networking – on a project that led to a strategic partnership with Ericsson for “mobile cloud acceleration solutions”. Again, it was too early – and the business model was wrong. Next, I consulted with IBM on their early work on edge computing in 3G and 4G architectures – still too early and 3G was certainly not the right place to start!
Meanwhile, in 2010 I’d written a White Paper [1] setting out my thoughts on the role of edge computing in generating new revenue streams to offset the race to the bottom in mobile tariff wars. When it was published, I got a cold call from the founder of a tiny start-up claiming to have implemented everything I was talking about in the paper…anyway, more about this later!
Another 2 years went by - but I had not given up on the edge! In December 2013 I ran a workshop with Intel and NSN (who subsequently dropped the “S” and re-birthed as Nokia Networks) on the outlook of potential use cases and deployment considerations for “Mobile Edge Cloud” based on 4G. This work led directly to the first edge cloud offering from a major vendor in the form of “Liquid Applications”. Definitely ahead of its time – but not in a good way! However, I learnt a lot!
These efforts within the ecosystem also had the beneficial effect of initiating a standardisation activity. In fact I had a place at the table in a meeting room in Bath, UK when Intel, Nokia and three or four Tier 1 telcos launched the ETSI MEC ISG (following in the footsteps of the ISG for Network Functions Virtualisation) and named the technology “Mobile Edge Computing” (the subsequent re-branding as “Multi-access Edge Computing” I had no part in!).
Coming back to the present day, six or seven years on, the truth is that in the consumer market we’re no further forward with the business case for a “deep” rollout of edge computing in telco networks (where “deep” means ~1000 telco aggregation sites in a large European market). Instead, most operators are barely getting their toes wet, ostensibly addressing “moderately low-latency” use cases with a “shallow edge” rollout (10-20 main switching sites in a large European market). This conservative approach will not enable the 5G low-latency vision. But operators are conservative by nature and like to see the cast-iron business case before placing their bets.
Meanwhile, we are right in the middle of the trough of disillusion with 5G. The benefit of hundreds of Mbps are not apparent to the user – and operators, governments and investors are getting distinctly nervous about the 5G business case [2], [3]. The killer application for 5G today is therefore not in the consumer space.
Where is the money (today)?
While this chicken-egg conundrum prevails on the consumer side, those hoping to find the elusive new 5G incremental revenue streams have turned their attention the the enterprise sector - and especially to Industry 4.0-minded enterprise customers who are interested in “cutting the cord” for all connected devices within their organisations.
Those with vested interests (and essential patents) in the 3GPP standardisation treadmill would have us wait another half a decade or so before the "ecosystem can be lined up" to follow their lead on implementation of URLLC techniques that are still on the standards bodies' drawing boards. This is a perspective loaded - as I said - with vested interests and is certainly unhealthy when those holding all the cards are not in any hurry to consider other options to accelerate time-to-market or leave a chink of light for new entrants.
I’m not personally interested in standardisation turf wars - especially when fought by patent trolls masquerading as technology companies. While 3GPP still commands the thrall of the industry, there was some excellent under-exploited work achieved in ETSI MEC that has been overshadowed by the hype around 5G Service Based Architecture. Like all standardisation rivalries they tend to stall, rather than stimulate innovation and adoption. I’m much more interested in harvesting the best thinking from both sides of the fence.
ETSI MEC’s fine-grain Traffic Offload Function (TOF) rule-base is certainly superior to the current level of standardisation of UPF capabilities in 3GPP. Another example is stateless and stateful application mobility between serving nodes at the edge. By decoupling these key functions logically from the 5G SBA User Plane Function, they become available today! No more waiting around for years until R16 and R17! And what’s more…they can actually be deployed in LTE as well as 5G networks! And this is music to the ears of those wanting to accelerate 5G incremental revenue streams.
Personally I’m a huge advocate of getting on with this! As long as we keep one eye on re-convergence of the standards with the ETSI and 3GPP functions co-existing symbiotically, we should definitely not postpone any longer the promise of ‘the Real 5G’!
Yet another logjam in 5G Private Networks – how to safeguard enterprise data sovereignty?
Contrary to received wisdom - and despite having dreamt for 10 years about heads-up displays with pothole detection alogrithms in AR cyclists' glasses - the "killer application" for MEC is not an application per se. There is a less obvious role for MEC in which it becomes the surprising logjam-busting hero in the story and it is this:
Restoring the data sovereignty and perimeter protection expected by any enterprise CISO who is embarking on building a 5G private network.How does this work and why does it need MEC? One incredibly powerful feature of MEC's Traffic Offload Function is that it enables a radical alternative to the complexities of putting a full LTE/5G network on premise for large enterprise customers in the age of Industry 4.0. The full EPC in 4G (or NGC/SBA in 5G) need not be deployed to each enterprise location – instead being much more efficiently located in public cloud or telco cloud.
But you do need a node on-prem to act as the perimeter off-net/on-net traffic steering function to safeguard enterprise data sovereignty. The best kept secret is that ETSI MEC TOF is actually the perfect solution for this! And before the 3GPP purists cry foul and tell me this is the job of a distributed UPF…I agree…but please hear me out…more on this later!
In performing this crucial role at the enterprise perimeter, a rich implementation of MEC is not only an enabler for low-latency applications, but also becomes an antidote for telecom operators to the new threat of fully private LTE/5G networks using unlicensed spectrum. This novel approach helps mobile network operators to remain relevant and to counteract the threat of network equipment vendors, system integrators etc selling private wireless network solutions direct to the enterprise market and cutting the telco out of the value chain. Similarly, it plays very nicely into the deployment of 5G mobile private networks over cloud service providers’ hybrid cloud architectures.
Who’s the approach and why is it a game-changer for Telcos and Hybrid Cloud Providers?
Here’s where I need to declare an interest. Full disclosure: I’m an Advisory Board member at Saguna [4]. That cold call in response to my 2010 White Paper was from Lior Fite, Saguna’s co-founder and CEO. “I saw your White Paper and wanted to let you know that we have solved all the problems you described and have a working prototype.”. I was deeply skeptical but Lior sent me a few patents and some documentation and as I began to read it became clear that he was right. That was the start of a 10-year journey that saw Saguna become the unsung heroes of mobile edge computing, contributing huge amounts to the ETSI MEC standards and enabling the fundamental research projects and early live network pilots for several of the largest telcos in North America and Europe.
Anyway, returning to the focus of this paper, Saguna call this alternative approach I’ve been describing to delivering managed 5G private networks “Hybrid Public-Private Network” (HPPN). HPPN allows the telco or cloud service provider to use its full portfolio of public spectrum assets to provide on-campus small cell coverage and capacity, while allowing the enterprise customer to keep selected cellular data within their own perimeter. The rules for which data stays inside and which data goes out to the operator’s core are extremely sophisticated and granular - down to individual IP flows or even to closed user groups of user equipments - while being easy to self-configure by the enterprise customer through a GUI or API.
MEC-enabled HPPN empowers the mobile network operator (or cloud provider) to offer a far more cost-efficient, secure and simple solution that meets all the enterprise customer’s needs for privacy and optimal routing of data directly to/from internal enterprise resources (Private Cloud, Intranet, IP-PBX, UCC) without them having to build and operate their own LTE/5G network. It also acts as a protective antidote to the threat of systems integrators and telecom equipment manufacturers disenfranchising operators and selling fully private mobile networks direct to the enterprise.
I have experience of this defensive strategy from the US market where a very large telco was able to offer a managed solution to a large semi-conductor manufacturer for “cutting the cord” in a large silicon fabrication plant which resulted in a 90% TCO saving compared with the dedicated mobile private network they had been offered from a Telecom Equipment Manufacturer!
What’s so unique about the approach I’m suggesting?
This novel and efficient approach adds an ETSI MEC-compliant TOF gateway into the customer premises to supplement the on-premise radio coverage solution. This comes at a fraction of the cost and complexity of heavier solutions that involve some or all of the core EPC/IMS to be replicated on the customer’s premises.
Whereas, most CUPS/UPF solutions today have a crude offloading capability whose decision-making is based on “Access Point Name” (APN) and therefore requires provisioning of settings onto users’ equipment (UE), in contrast Saguna’s ETSI MEC-compliant TOF implementation supports super-granular offloading rules (like a firewall’s rule-base) and requires no UE provisioning. Offloading is done at the granularity of users’ individual traffic flows and in/out routing decisions are triggered by URL, IP address, port number or IMSI user group. What’s more, because it is aware of traffic at user IP layer (not jut GTP layer), Saguna’s offering also includes a full DNS function and a DNS proxy, which cuts down the round-trip for DNS queries, accelerating name resolution for every flow, including those that leave the enterprise perimeter, thus providing yet another dimension to 5G performance enhancement.
But the Telcos' core network experts are committed to following the 3GPP standards, right? Well, fortunately, Saguna is also implementing a fully 3GPP-compliant distributed CUPS UPF in software, which will inherit all the aforementioned sophistication that other vendors’ solutions are lacking [5], [6]. What’s more, it will be cloud native from day 1, has been developed ground-up with performance and scalability in mind and will be implemented in tight synergy with all the above-described MEC functions. Only by taking this approach will microseconds be saved instead of added!
And what's more: the HPPN approach using MEC TOF is supremely pragmatic - it has far more flexibility to support a variety of architectures:
- Supports Industry 4.0 low latency use cases on existing 4G (LTE/LTE+) coverage as well as on 5GNR.
- Supports "Deep Edge" deployments at S1, N3 interfaces (using the ETSI MEC "bump in the wire" option [7]) allowing a MEC TOF breakout gateway and enforcement point to be deployed within the RAN/O-RAN architecture (rather than just at SGi or N6) - thereby placing perimeter enforcement exactly where it's needed for an enterprise customer.
- Supports far more sophisticated options for per-flow local breakout (5-tuple, HTTP etc) compared to the crude APN-based approach of a a typical EPC or UPF implementation.
- Requires no per-UE provisioning of APNs
- Supports all the architectural options in the "5G-ACIA Non-Public Networks for Industrial Scenarios" playbook [8]
- Supports seamless inter-UPF mobility with stateless or stateful application handover - important for large enterprise with multiple serving UPF nodes, either for scaleability or for 1+1 redundancy
The 5G LAN to 5G WAN up-sell opportunity
Once you have a footprint of these “dedicated” HPPN on-campus MEC/UPF deployments you can up-sell MEC hosted applications such as IoT security-as-a-service. You can also use the campus deployments as a stepping stone to wider MEC deployments in the macro RAN, by expanding a multi-tenanted HPPN MEC edge into the Macro RAN layer in the backhaul aggregation hub serving the cluster of macro cells surrounding these campus locations.
This way, the enterprise’s logical perimeter is still respected, even when the enterprise’s devices are moving between buildings (e.g. in the chemical plant, airport or hospital) and are moving between small cells and macro coverage layers. A side-benefit of this expansion into the Macro layer for the operator is a sure-fire business case for "deep-edge" MEC deployments in the RAN, based on solid revenues from the enterprise customer as an “anchor tenant”. Then, since it is multi-tenanted, the macro-aggregation-sited MEC node can be re-used for other enterprise customers in the vicinity or for consumer use cases for 5G and MEC. This will help to break the challenging chicken-egg problem of justifying widespread MEC rollout for the top-right quadrant low-latency use cases.
Unheralded? Not in leading 5G markets!
Saguna is making huge progress with this surprising application for MEC. For example, during 2019, AT&T piloted an enterprise solution based on Saguna's technology and has already made live deployments with Samsung Semiconductor, Boeing, Rush Hospital in Chicago and several other customers in different verticals.
In South Korea – arguably the most developed 5G market in the world – SK Telecom has selected Saguna’s vEdge+ software platform on HPE hardware as a key component in its 5G+MEC architecture, both for MNO network-based and for enterprise on-prem deployments.
Based on this real traction and the rise of CBRS-based private LTE/5G networks, I expect that this enterprise on-campus/dedicated use case for MEC/vUPF as an enterprise perimeter gateway will gain adoption much faster than the telco in-network deployments of MEC for the wide area and for the consumer.
Delivering on the real promise of 5G is a huge priority for those interested parties within the ecosystem concerned with finding a return on investment for the hundreds of $B invested in spectrum and infrastructure. An entire ecosystem is lining up – all wanting to deliver on that promise. But as we saw in recent weeks, one single obstacle can stand in the way of the entire 5G fleet of container ships [9]. What you need in this situation is a tiny, tenacious little tug-boat to come to the rescue!
References
[1] Early White Paper setting out the concept of edge computing in mobile networks. Predates the naming of edge cloud or edge computing by several years.[2] 5G from EE has arrived in my back garden. Now I just need to work out what I can actually do with 145Mbps. Streaming 30 HD movies simultaneously, perhaps? Not much use to me (one pair of eyeballs, one pair of ears). So who did the business case for bringing 5G to Abingdon? What is the ROI? When will the spectral efficiency gain pay back? I don’t need 145Mbps, let alone 1Gbps. Nobody does. But if they could get the latency down from 38ms to sub-10ms I can think of a bunch of useful things...
[3] In a follow-up to my most recent post about the lacklustre return on investment from 5G in China - the world's largest 5G market, here is a fascinating interview with Lou Jiwei, former Finance Minister of People's Republic of China.
[4] "Saguna's in-depth knowledge of the edge cloud computing industry meets and exceeds customer expectations. By offering clients game-changing solutions that provide unparalleled security, functionality, and reliability, Saguna is well-poised to increase its market share in the edge cloud computing industry.”
[5] Saguna Networks Ltd has announced the launch of ‘vEdge+‘, ...the first [product] of its kind to offer a UPF and MEC integrated data forwarding plane to meet the Ultra Reliable Low Latency Communication (URLLC) requirements of 5G networks.
[6] What’s the killer app for 5G? Please don’t show me another speed test or preach about spectral efficiency. Saguna Networks Ltd unveil vEdge+, combining all the MEC software components and the 5G UPF into a single software image, turning milliseconds into microseconds in order to deliver on the 5G promise of ultra-low latency for AR, VR and closed loop control for Industry 4.0 and C-V2X.
[7] ETSI White Paper: MEC deployments in 4G and evolution to 5G (Page 5)
[8] 5G-ACIA Non-public Networks for Industrial Scenarios
[9] Egypt's Suez Canal blocked by huge container ship
Add a Comment