Network Slicing

Almost every PowerPoint I’ve seen about 5G cellular networks talks about network slicing. This is a new networking term unique to 5G. This is the second article looking at new features of 5G, with the first being a blog on Massive MIMO.

Cellular networks are now expected to make multiple simultaneous connections with different characteristics. The examples used in most presentations explain how a cellular network should be able serve traditional cellular voice, more robust cellular data, IoT monitoring and connecting to self-driving cars. Each of these applications requires connections with different bandwidth, latency, security etc. The cellular network will be expected to immediately recognize the required need and respond appropriately.

It’s a challenge because of the diverse nature of each kind of network demand. For example, an IoT network will be comprised of huge numbers of devices mostly in fixed locations and requiring small bandwidth. Contrast this with cellular data, where as we increase data speeds we’ll expect the network to deliver large amounts of bursty bandwidth to mobile devices by combining multiple channels of frequency and even signals from multiple cell sites. Demands for self-driving cars or gaming will expect large and steady bandwidth with extremely low latency. These examples are some of the primary uses for a future cell site, but there are dozens of other kinds of connections that will be needed.

The ability to design a quick response to diverse network needs is made more difficult by the fact that every market for every cellular carrier uses a different combination of spectrum blocks and different channels within the blocks. This makes it impossible to design a ‘standard’ network strategy that will work everywhere. To be effective a cellular network must combine the spectrum components available in a given network to create a homogeneous network.

Landline networks are able to handle diverse types of demands using a combination of quality of service (QoS) and techniques like virtual private networks (VPNs). QoS uses a feature called differentiated services to classify and manage different types of IP traffic like streaming video, VoIP, web surfing, etc. Many networks also then use VPN functions like IP tunneling to isolate data paths aimed at specific customers.

These same techniques are hard to apply on a cellular network. Cellular systems need a sophisticated networking solution because the network is limited at any given time by the number of channels of frequency that are not being used. We don’t worry about this on landline networks because we can flood the network with enough bandwidth to accommodate every request. To most effectively use the available bandwidth a cellular network must quickly recognize the exact nature of the bandwidth being demanded and then cobble together the most efficient use of available spectrum. Current QoS solutions can’t adequately distinguish between different types of traffic to the degree needed to make this determination.

Network slicing provides a new way to partition the spectrum on a network. In layman’s terms it performs several functions that differ from QoS. Network slicing can quickly determine the nature of a bandwidth demand. It can then create a wide range of network responses.

One of the features of network slicing is that the network can be pre-configured for different uses. For example, a portion of the network can be isolated and assigned to a single function like IoT. Even more important, new revenues can be generated by partitioning and isolating a part of the network for a single customer – a business within range of a small cell site can be sold a share of the capacity of the cell site to guarantee better service. Slicing could also segregate traffic better – for instance, a cellular carrier could isolate traffic from one of it’s MVNO partners from other traffic on the cell site.

Network slicing can also subdivide spectrum. It allows the cell site to use a portion of a channel for a given connection rather than the whole channel. Slicing off small amounts of spectrum for small bandwidth needs if far more efficient than how cell sites operate today.

Finally, network slicing introduces a lot of new data features not available with QoS. The network can customize the way it handles any particular data stream in terms of data priority, encryption, data storage, etc. The network can more easily give priority to things like law enforcement connections, or IoT signals from critical devices.

Should All Bits be Treated the Same?

Polk County SignNetwork neutrality asks the question if different providers should be treated the same. That question asks if it’s okay to give priority to NetFlix over another movie provider like AmazonPrime. The concept has people up in arms because they understand that the deals made between content providers and network owners will end up restricting their choices. And people want choice.

But many of the articles I have seen talking about net neutrality have confused the issues between ISPs and content providers to somehow mean that there can’t be any discrimination among bits. And so I ask the question, should all bits be treated the same? And the obvious answer from a network engineering perspective is no, of course not. We already discriminate today for some bits and in the future it’s going to be desirable to discriminate a lot more.

Today any ISP delivering their own VoIP product already discriminates in favor of voice. Customers don’t want their phone call disconnected when another family member starts watching a movie or downloading a large data file. And so we give the voice packets first priority using techniques that are called Quality of Service (QoS).

QoS is the combination of a number of techniques that can give some packets better treatment than they would get using only best effort delivery. For example, QoS uses traffic shaping techniques like packet prioritization, application classification and queuing at congestion points to give priority to preferred bits. QoS also can use the Resource Reservation Protocol (RSVP) at gateways to fine tune the level of packet prioritization.

In the future there are going to be other kinds of packets that we will want to give top priority. Some things that come to mind are signals from burglar alarm, fire alarm and health monitors  which we will always want to have delivered as quickly as possible in case of the emergencies they were designed to monitor.

But we are also nearing a time when we are going to generate a lot of bits that we will want to give the lowest priority. We are going to have numerous monitors and sensors as part of the Internet of Things that will be delivering data that we will not want to interfere with voice calls or even video viewing or web browsing. It’s hard to imagine that we will insist on high priority treatment for packets from monitors that are looking at things like the humidity levels of a flower bed or the number of eggs left in the refrigerator.

And so I think it is likely that we are headed for a time when we will have three types of traffic in our homes. There will be the high priority packets for things like telephone calls and medical monitors. We will have regular priority for things like watching movies or browsing the Internet. And we will want the lowest priority for some of the background sensors that will keep an eye on our world.

And perhaps we will also someday get the flexibility for each household to choose which bits they want to give the highest and lowest priorities. It certainly is going to be a bit of a challenge for network operators, because the easiest thing to do is to treat all bits the same. But if the world demands different priorities for bits, then network operators will find a way to deliver.

What nobody wants is for our ISPs to dictate to us what we can watch by picking winners and losers among content providers. We want the option to watch movies from some start-up content provider and not be forced to watch NetFlix if they are the only ones with deep enough pockets to buy faster connections. If network providers take the path of picking Internet winners and losers they cannot be surprised when people flood to alternate network providers as they show up in any market.