For your average mobile subscriber circa 2022, 5G can seem like old news. After all, it’s been more than a year since the first 5G networks went live. Those of us in the telecommunications space, however, know we’re still at the beginning of 5G transformation. One could even argue that most of the steps taken so far have been merely a prelude to the real event. Now, with 5G standalone networks launching around the world, we can truly begin to reimagine telecom services. And the next big step on this journey is network slicing.
Since the earliest mobile networks, cellular services have been one-size-fits-all. Now, with network slicing, operators can carve out segments of their networks and tune them to the needs of specific applications. 3GPP has already defined the first three main slice types: enhanced mobile broadband (eMBB), ultra-reliable low-latency communications (URLLC), and massive machine-type communication (mMTC). Now, carriers can choose to use each of these slices – and likely many others – to offer customized services for enterprises, industry, the Internet of Things (IoT), and much more.
Despite the huge impact slicing will have on telecom services, however, many in the industry still don’t fully grasp how it works. That’s understandable; the topic gets very technical quickly. But don’t panic. Even if you’re not versed on the latest 5G standards, we can help walk you through the basics.
Back in 2019, HPE Distinguished Technologist Marie-Paule Odini wrote a series of blogs on the topic. The series continues to be among the most well-read at HPE.com. If you’re looking for just the highlights, here’s an overview.
Understanding slicing
Blog 1 starts by asking just what network slicing is and why it matters.
In fact, slicing was one of the biggest drivers for 5G. Industry groups recognized the growing need for operators to offer services tailored to the requirements of different customers and applications over a single infrastructure (Figure 1). Some services, like mobile broadband, can serve a variety of use cases. Others, like highly secure services for government, or low-latency services for remote surgery, must meet more specialized requirements.
To meet this need, the Next Generation Mobile Alliance (NGMN) introduced the concept of network slicing in 2016. Odini helped explain the concept in a paper published that year, which defined a set of devices and applications with a number of policies around bandwidth, latency, geographic coverage and capabilities, and other attributes.
Odini explained how an end-to-end network slice instance (E2E-NSI) could be deployed, instantiating a combination of access, as well as fixed and mobile network resources across transport and core subnetworks. These subnetworks can be composed of physical network functions (PNFs) and/or virtualized network functions (VNFs), with an information model defining the structure of the components.
She then ran through the components of a typical 5G network architecture, noting that slicing must be end-to-end, including the radio access (RAN) and transport network, but also crossing multi-access, multi-domain networks, and potentially even multi-operator domains. Odini laid out examples of the various network functions participating in a 5G slice (Figure 2), which would be placed at the edge or core, depending on policy or latency requirements.
Finally, the blog discussed the many different use cases possible for network slices, offering an example of how one customized slice—for connected vehicles—could touch on multiple slice types (Figure 3).
For more detail, read Blog 1.
Network slice management
Blog 2 dives into the management of network slices.
It begins with the specifications defined by the European Telecommunications Standards Institute (ETSI) and 3GPP to integrate with the ETSI Network Functions Virtualization Management and Orchestration (ETSI NFV MANO) framework (Figure 4).
The specifications introduce three new management functions for network slicing:
- Communication Service Management Function (CSMF), which relates to the customer service
- Network Slice Management Function (NSMF), which deals with the end-to-end slice
- Network Slice Subnet Management Function (NSSMF), which deals with the underlying network.
While NSMF and NSSMF request some subnetworks that translate into NFV MANO Network Services requests, slices typically span multiple domains, such as heterogeneous access including RAN, transport, edge and core. Operators will use multiple NSSMF instances to manage the underlying subnetworks. For example, the RAN NSSMF may interface with RAN controllers to request RAN resources, while the transport NSSMF interfaces with a software-defined network (SDN) orchestrator to request certain links (Figure 5).
Each domain can have a mix of physical and virtual functions and resources, as described in a hybrid inventory that’s automatically updated with new instances and relationships. Alarms and events are collected and processed by service assurance with a combination of topology correlation and artificial intelligence that identifies patterns.
The typical lifecycle management process entails:
- Design and preparation of a slice template
- Instantiation request to create, configure, and activate the slice
- Monitoring and control of the slice to meet quality-of-service (QoS) requirements
- Deactivation once the slice is no longer required.
This process is handled by the E2E slice manager, which transmits interactions (request/response or subscribe/notify notifications) to different tenants northbound or to underlying management systems, such as NFV MANO. Since slicing happens across multiple domains, it cascades down to the resource level to ensure proper traffic isolation, bandwidth or latency. This involves a dynamic topology, so slice management requires very efficient programmatic end-to-end orchestration, with closed-loop automation across fulfillment, inventory updates and service assurance.
Finally, Blog 2 details HPE’s history in service provider orchestration and introduces our Telecom Operations and Network Automation portfolio to enable it, including HPE Service Director and HPE NFV Director (and now, HPE 5G Automated Assurance).
For more details, read Blog 2.
How many slices do you need?
Blog 3 discusses the productization of network slices, asking how many slices actually make sense.
This is not a simple question. When combining the three initial 3GPP slice types with other customizable settings, operators could in theory create up to a million distinct slices that devices could recognize. Of course, to sell those slices, operators need to map them to actual customer-facing services that can be catalogued with a specific service-level agreement (SLA) and price (Figure 6).
It can be instructive to return to the connected vehicle use case described in Blog 1, which demonstrates how complicated this question can get (Figure 7). Should an operator launching a connected vehicle service just use the three main slice types (eMBB, mMTC and uRRLC) or create additional ones? What about designating slices for different auto manufacturers and models? Operators will have to find the right balance between meeting more specific needs and minimizing complexity. And they’ll need to do it on a case-by-case basis.
For the full discussion, see Blog 3.
5G RAN Slicing
Blog 4 dives into the technical details and challenges of implementing standards-based network slicing across a multivendor disaggregated RAN. In a 5G network, any access network should provide a standardized mechanism to define and operate network slices, using open APIs to support access networks that are multi-vendor, disaggregated, virtualized and distributed. That includes the RAN. However, at the time the blog was written, it was still not clear how exactly this would be done.
Odini offers a comprehensive technical discussion of the topic, walking through a hypothetical deployment for an operator creating slices for fixed wireless access, mobile broadband for indoor and outdoor venues, IoT and connected vehicles. Odini discusses:
- Cataloging technical details for defining RAN subnet slices for each proposed service
- Stitching together RAN resources and associating them to subnet slices, including positioning DU and CU components
- Deploying RAN slices, including how operators might choose between the different RAN split options
- Modeling the RAN, including understanding the number of interfaces each RAN entity will have with virtual links, and the specifics of those links (bandwidth, latency, traffic type, etc.) to create the 5G RAN information model
- Determining how the slicing information model interacts with the 5G RAN information model
- Mapping resources to slices for each of the different slice types the hypothetical operator is deploying
- Placement of the different network functions, depending on slice type and policy.
Ultimately, operators need a flexible slice management and orchestration system to be able to cope with the diversity of underlying RAN technology options, as well as RAN controller technologies, SDN networks, NFV MANO and other management systems.
For the full, in-depth story, read Blog 4.
Learn more
Today, service providers and vendors are taking concrete steps to bring network slicing to real-world networks and customers. As we do, we can redefine what telecom networks can do for enterprises and industry, and help service providers launch new business models that weren’t possible before.
Keep an eye out for new service provider offerings that use network slicing to address diverse enterprise use cases. And in the meantime, to brush up on the technical underpinnings of this transformative technology, check out our 5G news and read HPE’s complete network slicing blog series.
Oded
Oded Ringer leads Telecom portfolio messaging in HPE’s Communications Technology Group (CTG). He is responsible for bringing together products, business, operations, sales, and marketing into a coherent go-to-market strategy across all regions, countries, and market segments.