Multi-service Networks: Convergence and Adaptability

Modern networks are continuing to converge on a single physical infrastructure and protocol set, with the data network integrating applications which previously required their own dedicated networks.  Much of this can be attributed to the adaptability of Ethernet and IP, which have been able to evolve from their origins as ‘data-only’ to supporting truly multi-service networks through increasing speeds without greatly modifying their original functionality.

This is visible in the modern home network. 10 years ago, most homes had 3 types of network connections:

1. The voice network

              The plain old telephone system that has been a staple of the vast majority of homes for decades, connecting telephone handsets to the phone network with dated copper cabling.

2. The video network

              Cable or satellite video networks such as Comcast cable or DirecTV, using their own dedicated physical infrastructure – either coaxial cable to the home, or a satellite dish and receiver used to receive only video content and related limited-use data such as electronic program guides.

3. The separate data network

              Dial-up, ISDN or later broadband internet connections, whether wired or wireless. Although this network may share some physical infrastructure with the voice or video network depending on the technology used, it is billed, managed and operated separately, and provides very different capabilities.

The trend of ‘cord cutting’ (or, in other terms, removing unnecessary redundant network connections to the home) is seeing the separate data network taking over the tasks that once required the dedicated voice and video networks in the home. As its capabilities have increased, the data network has not only been able to provide equivalent service to those dedicated networks, but has actually surpassed them.

As well as providing higher-quality versions of traditional voice and video services, the data network has also made possible newer, higher-value approaches to the problems of voice connectivity and video delivery: systems like Skype and FaceTime that incorporate much more than simple voice calling and on-demand streaming services such as Netflix and YouTube are uniquely enabled by the capabilities of the data network, and as such have greatly improved its value when compared to the separate dedicated networks previously used for these needs.

What were once entirely separate networks and pieces of hardware have really been replaced with software, and are now operating as services on top of a converged data network. This is a huge change to the network, media and telecommunications landscape in a fairly short amount of time, and many incumbents are still reeling from its continued effects.

In the enterprise, a similar story is playing out. Dedicated voice systems with expensive PBXs are largely being replaced by soft-phone VoIP solutions, either locally or cloud-hosted, that offer more integration with other collaboration systems and (hopefully) a lighter monthly bill without the need for local voice hardware. Some enterprises who invested in high-end videoconferencing systems also had dedicated video networks of a sort, but they were typically dedicated data network connections for the video traffic rather than a cable or satellite video delivery network connection as typical in the home.

However, the enterprise did – and still does in many places – have a fourth type of network:

4. The storage network

              Whether using dedicated Ethernet infrastructure or a purpose-built protocol and physical layer such as InfiniBand, storage networks attempt to provide zero-loss and consistent latency to their applications, which are connecting servers to storage or interconnecting storage arrays.

This is largely because the protocols which handle storage data transfer such as iSCSI are not designed to handle lost data in the way that internet applications using TCP/IP are; they are extensions of disk-to-system protocols that run over internal system cabling less than a foot in length, on a dedicated connection, rather than contested networks serving many applications and nodes.

Whether a similar story of convergence when it comes to network storage will occur is an interesting question, and not one that I will endeavour to answer here. However, the costs of building and maintaining entire separate networks for storage are considerable, and many organisations would dearly love to invest in only a single network.

What the convergence we have seen to date illustrates is that building an adaptable network is not only a sensible choice for the needs that we can see today, but it also gives us the best chance at reacting as quick as possible to the needs we don’t know are coming. This adaptability comes in three main factors:

1. Technology

              Ultimately, if the technology chosen for the network is simply not capable of supporting new services, no matter how much headroom it has or how well-designed the network is, it will be unable to keep up with the needs of new use cases and applications.

2. Headroom

              A technology can be suitable for carrying different types of traffic, but lack the technological roadmap to economically improve throughput and latency to a level where supporting these different types of traffic and the applications that create them is really feasible on the needed scale.

3. Design

              All the technology in the world can’t save a poorly-designed network. Although it’s very difficult to design for use cases which may not exist yet, taking a scalable, adaptable approach to design from the beginning makes it a lot easier.

Enterprise and home networks are similar and dissimilar in many ways. One thing is for sure – they both stand to benefit from and be impacted by the continued convergence that we will see in the next few years, and to support this both need to be as adaptable as possible, using technologies which are fit for the purpose.

How much further will Ethernet and IP be able to take us with ever-more converging networks? As it stands today, they are not the bottleneck – partly because applications have been developed to get around their idiosyncrasies, and partly because they have proven to be extraordinarily adaptable. For some time to come, they will remain the bedrock of our networks, whether enterprise or home.