Much Ado About Slicing
Who needs network slicing and why?
To say much has been said and discussed about the 5G business case is an understatement. In fact, it seems that the one thing the industry agrees upon is the fact that the 5G business case is questionable, at best. There is no doubt that early adopters may be willing to pay more for 5G, but the vast majority agree that consumers will pay only 5% more for 5G, than they currently pay for a 4G package. Moreover, most agree it will take operators years to break even on their 5G investments. So, where is the money?
The first, obvious answer is that the money is in entering new/additional markets. Some mobile operators talk about entering the cable/video streaming market. Others about improving coverage via Fixed Wireless Access in some rural areas where they currently have little business. And, indeed, these first 5G services will cover some of the costs incurred.
The second, perhaps more future-thinking line of thought, suggests that carriers will be able to launch a wide range of new services with 5G, services that will offer customers unique value. We are talking about services with unique requirements, SLAs and profiles. Services that do not exist today. These tiered services will be priced differently and therefore offer a new source of revenue. And yet achieving these new, tiered services with guaranteed SLAs will require a change in the way networks are architected. Or will they?
Of course, those of us who are not that sure about the launch of these ‘brand new’ services - which may or may not be put into practice – prefer to continue working with the network as it is today. I mean, why fix that which is not broken? When 5G drives more traffic, it’s best just to ‘throw more bandwidth’ at it and things will work themselves out. No?
The answer is yes and no. 😀
Yes, the network does a pretty good job today, but let’s face it, massive overprovisioning and ‘good old college try’ approach to guaranteed SLAs has already been proven as severely limiting. As we have discussed before, most networks do not have the measurement and telemetry systems in place to proactively guarantee SLAs.
In comes slicing
Many years ago carriers built different networks for the different services they offered. A voice network, a data network, a synch or OT network and so forth. But decreasing margins and the need for reducing CAPEX have forced them to rethink their network architecture. In the 5G world, building separate networks for each type of service is just not viable. So network architects and organizations (IETF, ITP, and others) have come up with a solution in the form of network slicing.
Network slicing is an architectural concept that allows the cutting of one physical infrastructure into multiple virtual networks. Each slice is then customized to meet the specific needs of the applications, services and resources that need to run over it. In this sliced world, carriers will build a super high capacity network, and then allocate resources to optimize and prioritize the traffic which runs over it. Some of my colleagues have suggested that network slicing is only an evolution of today’s networking technologies spiced up with some of the newer capabilities the industry has been discussing for years.
And while it may be true, slicing is more complex. The industry is differentiating between ‘hard slicing’ and ‘soft slicing’. To simplify, ‘hard slicing’ refers to the provisioning of resources in such a way that they are dedicated to a specific service or slice. Whereas ‘soft slicing’ is defining the VPN path based on KPI constraints (PCE based on performance monitoring). This, of course, suggests that soft slicing cannot absolutely guarantee the availability of resources nor the SLA. The difference between enhanced VPN, Segment Routing and Hierarchical QoS, is the methodology which defines the slicing structure. The only way to ‘absolutely, positively’ assure that a slice receives ‘X’ network resources is to hard slice it.
The fact is that today we have a variety of technologies with which to slice the network. And industry experts agree that a combination of these technologies is the best way to move forward. These technologies differ in their level of isolation as well as their cost-efficiency. Obviously dedicating resources is more expensive/less cost efficient than sharing resources. The following chart helps visualize:
Who needs hard slicing anyway?
I admit, I have heard from customers that the benefits of network slicing do not outweigh the complexities (and cost) of re-architecting the network. And while that may be true, our engagements have demonstrated 3 distinct use cases where ‘hard slicing/isolation’ is being asked for:
Hard slice per service class (SST/S-NSSAI for eMBB, URLLC, MIoT etc.) as per release 15
The industry defines at least three distinct classes of service brought on by 5G. Each of these classes, on a high level, is different than the others in the type of resources it requires and the accompanying SLAs. Enhanced broadband, for example, requires a lot of capacity but not exceptionally low latency. Whereas, for ultra-reliable low-latency (URLLC) services, resources need to be allocated in a way that provides a highly reliable and resilient service with stricter SLAs .
Use Case #1: Slice per service class
Hard isolation per carrier or MNO
It seems that the industry has been talking for years about shared network resources. However, 5G may be the turning point. The higher 5G spectrum frequencies travel much shorter distances than their 3G/4G predecessors. This means that cell-sites or antennas need to be installed much more closely together, especially in city-centers. If in the past, network sharing could reduce TCO by 30%, it is believed that in the 5G era TCO can be reduced even more. In the case of network sharing, the only way to absolutely, positively assure that one carrier is not discriminated against by various soft slicing measures is to implement hard slicing which dedicates resources to each carrier.
Use Case #2: Slice per carrier/MNO
Hard isolation for RAN and NGN co-existence
One last, and perhaps less obvious need, is to provide better or different SLAs to services with different ARPUs. In the case where the MNO intends to provide both 5G internet and broadband internet over the same network, they are likely to prioritize the 5G traffic over the general broadband traffic. By allocating the resources for the different types of traffic, these operators will be able to guarantee SLAs, and of course prioritize their more profitable 5G services over less profitable previous generations.
Use Case #3: Slice per ARPU
To be honest, we do not see one single killer application or use case for slicing. We often see a combination of the three use cases described above. Granted, as one of the few vendors in the world who offer a wide variety of both hard and soft slicing and isolation technologies, it might be said that we are biased as to the need for network slicing.
Of course, I cannot decide for you whether you do or do not implement slicing in your network. What I can ‘absolutely, positively’ say is that while you may not know today if you will need slicing or which slicing technology you will need, chances are you will know it only in the future. With that in mind, as you upgrade your network (and network upgrades are a day by day phenomenon), shouldn’t you look for solutions giving you the freedom of choice? If you do not, then it is more likely that your next network upgrade will come faster than you hope.
To learn more about ECI’s 5G Solutions, click here.