Recently, experts from VoltDB and Openet teamed up to co-host the webinar, “Monetizing Enterprise 5G: Harnessing the $600B Opportunity with Real-Time Data”. The webinar focused on the lucrative opportunities of a 5G-enabled enterprise for telcos and why the ability to harness real-time data is an absolute necessity in empowering the full monetization of 5G-enabled services.
Attendees asked a number of excellent questions during the presentation and we’ve highlighted those questions and their answers below. For those who missed the webinar, you may view it on demand.
Editor’s Note: Unless otherwise noted, the question is coming from a webinar attendee via our chat functionality during the session. Answers were provided by VoltDB’s Chief Product Officer, Dheeraj Remella and Martin Morgan, VP of Marketing at Openet. These answers have been edited for clarity and grammar.
Q: Could you explain a bit more about the role of machine learning in real-time decisions?
Dheeraj Remella (DR): When you look at machine learning, there are two roles – actually, multiple roles but two primary roles. One is being able to learn from the data and tell you what needs to be done. The second role is being able to refine what it just told you, so it’s basically self-correcting behavior that is essentially going to evolve what you know from your data so you are able to do better and intelligent decisions and avoid your false positive central, so this is really important. Machine learning is only useful when you take that inside and deploy it into a platform that is making closer to real-time decisions based on what it learned. If not, it’s just a very academic exercise at that point.
Q: How can the issue of heterogeneous migration be addressed?
DR: It’s really interesting because one of the things that you would see is that there are databases for specific functionality but at the end of the day, it is like a single system that you will be looking at, so that’s one point of view. The second point of view is don’t think of it as heterogeneous database but rather ingesting heterogeneous formats of data. So, when you have several streams of heterogeneous formats of data coming in, being able to intercept these streams and collate and correlate them is a really, really important capability that you would require in an environment where you have a variety of sources sending data in a variety of formats.
We have examples of that nature within VoltDB where we have sent the smart meter data coming in from a variety of sources and we are able to collate that all together, make sense of it, arrange it on timestamp or however it be your business rules that would require you to arrange that data in order so you can create this environment of a complex event that it’s going to trigger an action. That would be my approach. Instead of thinking of a database as a place to store data, start thinking of a data platform that needs to act on data as it is arriving; so when you can actually convert your point of view of what needs to be done in those streaming environments, you will see that it’s a much better scalable point of view and you can solve your problems differently. Instead of saying, “Why is this not working,” think of, “How can I make this work?” When you do that, the heterogeneous nature becomes just a factor that you can solve just by applying converters or adaptors, kind of architectural pattern where you intercept the data and homogenize it to what you need to do.
Q: Will enterprises pay a premium for 5G?
Martin Morgan (MM): That’s a good question. I don’t think we would pay just for 5G versus 4G and I think that’s kind of missing the point. I think what they will pay for is service-level agreements and guaranteed quality of service. When you’re looking at critical IoT applications, critical communications, for example, high-end security, to be ensured that these network levels of latency and quality of service as expected, then yes, enterprises would pay a premium. But if it’s just, “Hey, 5G is faster than 4G and we expect you to pay a premium,” that could last about five minutes. I don’t think you’d be able to charge for that but if you actually charge based on the service-level agreement for the different attributes of the network choices, then yes, I think enterprises would pay a premium for that.
Q: How can 5G and internet of things affect the healthcare system?
DR: The way I think of it is the healthcare system have two types. One is critical healthcare, another is non-critical healthcare. Non-critical healthcare, meaning like you have your arrhythmia monitoring patch on the chest and it’s sending information then you are able to actually detect what is happening. Actually, that is a critical part, sorry. I’ll swap the example. The heart arrhythmia monitoring is the critical one where it’s life or death kind of a thing. The real-timeliness, and that cannot be over-emphasized at all because you need to be able to have devices that a patient can use to resuscitate themselves and be trained in that kind of a scenario where the notification is immediate or someone, not the patient himself of course, someone close to them that are co-residing, say an in-house nurse or someone, but on the other hand, you can have a non-critical notification like, “Hey, you missed your medication,” and if the box did not open at a certain time, then you can actually get a notification that like, “Hey, you forgot to take your medication. It’s time. Go find it and take it.” These are the two systems that I can look at. There are so many metrics that the body is generating. You can think of your fitness tracking watches. Just from your blood flow and body temperature, they are able to say a lot of things. These are health-related, how to promote healthy living but if you flip it on the head and say, “How can I actually use this information to predict what could potentially go wrong?” Those are the opportunities that are going to start opening up because the capability is there. Definitely, medical research, that would be facilitated by such capabilities.
MM: I think healthcare is a very interesting example both for consumer and for enterprises. The healthcare example Dheeraj gave, the basis of watches, monitors, whatever, creating human data, and you take feeding it back to the GP or the insurance or health provider, so the fact where people could have a health checkup while out on a jog or a run. Maybe insurance companies should pay for that so they encourage people to give away a fact of whatever and people are encouraged to take exercises for the premiums to go down, the cost to the health insurance go down as well, but also in hospitals and healthcare providers, you’re meshing critical services that are dependent on very good communications. If the Wi-Fi is not up to scratch, this was a service-level agreement, so the Wi-Fi provider can operate or start to sell individual network services for critical monitoring devices, critical communications for healthcare professionals and for patients’ sources and charging service-level agreements to the hospital as need be so these are opportunities for telcos to get all your critical needs, critical machine needs, critical communication needs for a hospital. We can sell you these and you have a guaranteed service-level agreement which we will monitor and implement. That way, the host would probably pay a premium for that rather than just wait for Wi-Fi per month or 5G per month so the opportunity for both in consumers and in enterprises for healthcare.
Q: With 4G still being far off in some Asian countries, how do you see them getting to 5G in the future?
MM: 5G is a big investment, there’s no doubt about that and in order to get there, you really got to stop monetizing the existing 3G and 4G networks. I’ve seen a lot of creativity in some African countries about how they are monetizing 3G and 4G services. You’ll see for example, Vodacom in South Africa, just I think it was two months ago, they have launched free internet on 4G but it’s ad-funded so people watch adverts and the advertiser will get so many gigs of data per month if they watch advertising, and the advertisers pay the mobile operator. That’s a new business model. People start to consume data over 4G but they’re not paying for it because it’s ad-funded, and there are other ways, too. For example, governments are starting to work at using 4G to disseminate a public information, healthcare information, education, and the governments will be subsidizing those services as well so you could possibly say, “Yes, that is building the business case for 5G investment in the future in a lot of these countries where they are still building up the penetration of 4G services, and is looking at different business models. It’s not just looking at the end consumer cases, working out if the government subsidizes it for healthcare for education, these public services and also possibly looking at advertising, subsidizing from services as well, so it’s those new business models that can be leveraged rather than just saying, “Yes, the consumer must pay for everything,” because in some countries, to build 5G networks in the back, all the consumer pays for everything, it could be a tough argument.
Q: Does VoltDB integrate with message queues, and perhaps you could also say how?
DR: Yes, absolutely. I think I can actually see where this question is coming from because a lot of the network core, 5G standalone network core at least, if you look at it, is granted be fully microservice architecture and you have multiple ways of communicating within microservices which is one is a common storage that provides the same set of version of truth to all the services while also you need to be able to move control from one microservice to another which is where the microservice message queue would actually play a role. Volt, towards that end, actually integrates with several message queues and you can also create custom message queue integration if you prefer to depending on your choices but it integrates frontend and backend, so we try to bring together the storage and the Kanban style system where, Kanban is like a signaling system for taking action as the next step so we accommodate that model with the way that we are architected so it’s perfect for a good microservices architecture implementation.
Q: Can users do slicing in 5G NSA? If so, what changes are required in existing EPC to achieve it?
DR: One of the things that you would see with the change between EPC and 5G SA core is essentially, EPC has a variety of databases so with 5G SA Core, the approach is to unify the data and create like a unified data management layer, under which, you have all of you subscriber data or things data within the UDR, which is the unified data repository and the UDSF which is the unstructured data storage functions. When you look at what purpose these two serve, then you can actually start mimicking it. If you look at the purpose, the UDR is to provide a singular interface through like a UDM frontend for all the microservices to be able to interact with the data without touching the data so to speak, so it will actually interact with known interfaces. You have the same kind of a capacity in EPC as well. In EPC, each function takes any P Gateway or a PCRF, any of those are interacting with each other in known interfaces. Now instead of creating a point-to-point interface, you can mimic that same capacity of the unified data management frontend within your EPC so that’s an additional functionality but that moves your EPC closer towards a smoother migration towards the 5G standalone core. While your infrastructure is NSA, you can start evolving your core from EPC into a 5G SA state by starting to implement these frontend interfaces so you don’t need to rely on point databases, and then once you have this interface, you can always consolidate data later on and just change how these interfaces work to point towards this consolidated UDR and UDSF model instead of going to independent databases like HSS, HLR, and things like that. I’m hoping that answers the technical parts of it. From the slicing standpoint itself, you can slice the core part of it but if you haven’t moved to vRAN, that part and the session border controller, you’re going to have challenges there from slice standpoint but you can implement it by how 4G actually – if you look at roaming, so if you’re doing international roaming, my mobile phone is going to tell the RAN that like, “Hey, this is my network ID or network name,” so using that name, there are different sets of rules of how I get service so you can tap into that model in an NSA environment to say like, “Hey, if I actually have a set of things that belong to a certain application, they can all actually come in and semantically tell the network, the SBC that like, “Hey, I belong to this application and that can actually translate into QoS and PCRF rules that need to be in play based on the application name so it’s mimicking. That’s the way I can actually think of it, not like truly what network slicing would be but like mimicking those capabilities and building towards a smoother transition towards a 5G SA core and a proper network slicing. I hope that answers the question. If I don’t make sense, let’s actually have a conversation so I understand better.
Q: Can you explain the need for all the capabilities listed in the holistic approach for every application?
DR: What I actually say with those nine things that I actually mentioned about the holistic approach and the nine things that you need to do with your fast data, it might not be necessary for everybody. If you’re just taking – I’m going to veer away from telco for a bit. Think of clickstream. So, clickstream, do you need to act on all of it before you understand what you need to do with that data? Probably not. At that stage, a simple collection point and being able to migrate it to an analysis later, it’s just you don’t need to make real-time decisions because you don’t know what decision you need to make but once you actually have the intelligence to create a connected environment especially in telco and industrial IoT, there are well-established, well-documented processes. It’s not an agile environment where anything could generate anything in an insecure manner and anything could happen. That’s not it, right? You have a very well-defined process, well-defined outcome set and you want your production to be within a set of guardrails. In those kind of scenarios, you would see that all nine of them are important. There are several applications and scenarios where the criticality is way lower and the latency tolerance is way higher. In those cases, a stacked approach could just work fine because you could probably find free software for each later and you might even be able to do it at zero cost other than the developer and maintenance and compatibility problems but software cost-wise, it’s zero cost. I hope that helps.