Livin’ on the edge


This article was published in The Tech Capital on June 14, 2021. You can read the original article here

Why telco operators need to implement disaggregation at the edge


Written by Hannes Gredler, founder and CTO of RtBrick

Operators who would like to follow a more future-oriented approach should reconfigure their network systems based on a cloud-native approach. 

The telco market is in a state of flux. This year has brought about a global pandemic that has radically shifted consumer behaviour and challenged operators’ future-ready and efficient network infrastructure perspectives. With so much demand from consumers, it is natural for telco networks to become strained.

Coupled with that, the boundaries between cloud services and telecommunication networks are beginning to blur, as customers demand better performance across internet, public, private and hybrid clouds. Computational and transport latency, i.e. the time it takes to move packets (data) between the client and the server, have also been pushed to their limits, leading telco operators to rethink the way they deploy and manage their infrastructure.

The network edge

The network edge does exactly what is says on the tin. It migrates compute resources to the edge of networks by reducing the physical proximity between the user and resources to distribute workloads for faster performance. The goal here is to eliminate latency or data congestion issues and improve application performance. It also has the potential to improve network security, help lower costs and increase scalability.

By definition the ‘network edge’ refers specifically to edge compute locations, which are sites or points of presence (PoPs) owned by a telecoms operator. This includes, for example, a central office in the mobile network or at an ISP’s node. Another example would be the regular user streaming their favourite TV show on Netflix in HD. He or she won’t be concerned with the tenth of a second it takes to load up the content, however, for an operator, there are certain applications that need to handle capacity and process the data load at certain rates, and these will scale better if they are distributed to the network edge.   When it comes to managing edge computing technologies and their underlying infrastructure, the telco industry is a special case. Some of the world’s largest network operators are beginning to displace parts of their network functions closer to the edge, with the goal of creating new vertical business opportunities, shortening provisioning, and making networks more agile. However, telco operators are not only responsible for their own network infrastructures but also their customers’.

Even though operators are confronted with an inexorable growth in data traffic, consumers are still not willing to pay more for services. In fact, they can blame poor service and poor quality solely on the operator, when it is the source content that is at fault. The main problem here is that many operators are still operating on legacy infrastructures, which limits flexibility and elevates costs. For example, device implementation and maintenance are still time and cost intensive because workers are often required to manually configure individual systems. Anything that involves manual processes can exaggerate overhead costs, and can become cumbersome over time, leading to potential human errors and issues with the network. So, what are the options out there that will improve your network?

Disaggregation at the edge

The two main issues driving operators towards web-scale solutions are the costs associated with high spectrum licensing and higher operating costs and the length of time to deploy new infrastructures. Waiting years to increase the amount of infrastructure in a network that can’t react to unanticipated traffic demands is simply not feasible anymore. The simple solution mirrors how the giant ‘cloud-native’ companies have operated. Operators who would like to follow a more future-oriented approach should reconfigure their network systems based on a cloud-native approach, by disaggregating their software and hardware. These systems are not only more cost-efficient, they also allow for zero-touch provisioning and highly granular remote management.

Disaggregation means deploying network software separately from the hardware. This enables the case for using ‘merchant silicon’, which offers the same capabilities on high-volume, low-cost networking chips that the traditional network equipment vendors used to have in their customised systems. It allows operators to build networks from a new category of powerful low-cost ‘bare-metal’ switches that can then be turned into IP/MPLS switches.

Both hardware and software can be deployed using zero touch provisioning, without human configuration. Once installed, operations staff can deal with a single homogeneous equipment and operating environment, regardless of the software running on it, instead of needing to be trained in multiple vendors’ systems and processes. They can upgrade the capacity of any dimension of the system, within minutes, without throwing away the existing infrastructure.

By adding dedicated API’s, which allow operators to control individual switches as well as entire networks, with REST-based interfaces enabling the management of different types of systems and equipment from multiple vendors, the entire system becomes automated and future-proofed. This will give operators a myriad of options, including programming individual switches or entire networks, accessing underlying systems with complete granularity, unifying network and IT operations via a consistent REST-based approach ,and operating disaggregated components alongside other equipment using the same higher-level systems.

In addition, there’s also the access to a community of open Web 2.0 tools to add even more functionality, that comes part of deploying web-scale solutions. This includes:

  • Grafana – a visualization tool
  • Graylog – a log management system
  • Prometheus – a monitoring tool
  • SNMP and GNMi
  • RBMS API – exposes any of the RBMS GUIs to higher-level systems

With the individual needs of customers growing in complexity network operators will encounter special use cases, where off-the-shelf controllers are not desirable. Thus, management teams need to become proficient in developing their own in-house controllers in order to customise infrastructures to their own requirements, rather than being forced into a service model by their vendors. In the end this means that there is no cookie-cutter solution to best manage telco edge infrastructures, as there will always be differences and individualities to each system. Therefore, the tools and solutions network management teams use not only need to meet the basic needs of the operator, but also be highly flexible, compatible, and open.