Edge virtualization

Edge virtualization is the practice of using software versions of physical computing resources at the edge of a network, closest to the devices that produce data. In virtualization, the entire software stack -- including operating systems (OSes) and everything that runs on them -- is separated from the underlying hardware. Instances can then be copied and distributed to many different types of hardware. This is valuable for at the edge because the hardware at the edge where the environment is involves limited bandwidth and latency issues, is varied and dispersed geographically, and needs to be managed independently from the geographically distant core data center.

Edge virtualization is important because it extends the software-defined concept of the cloud universally. This software-defined approach enables the remote provisioning, management and monitoring of edge devices across large geographical footprints, providing a more secure and cost-effective alternative.

Edge virtualization, by its very nature, is a combination of software-defined compute, storage and networking -- much as you would find in the cloud, but where these resources are remote and, in most cases, of modest scale.

Where the cloud has massive compute, storage and networking resources that are tightly geographically associated, usually in one or two very large data centers, edge virtualization has a very large number of locations, each with a modest complement of resources.

Edge virtualization extends the fabric of compute, storage and networking to many remote locations, enabling the processing of a wide range of workloads in places where they need to execute.

In retail and hospitality, edge virtualization is rapidly becoming a key requirement due to enhanced levels of customer experiences driving the need for local processing. Applications such as point of sale (POS), loyalty, kitchen management systems and RFID (radio frequency identification) require faster local processing.

Virtualization's role in edge computing

Edge computing is the term that describes the duties of devices at the edge of the network. Every network has a core and an edge, a trunk and branches. The core (trunk) is where the network starts, and the edge (branch) is where it ends. The edge is the point at which external users interact with the network. These end users may be actual people or other external devices known as peripherals.

As the number of endpoints at the edge increases and become more varied and dynamic, it becomes less efficient for each endpoint to reach back to the core for resources at every external request. This is where virtualization comes in.

Virtualization puts resources at the edge. It allocates virtual resources to the endpoints -- or branches -- so that instead of constantly requesting resources from the core, endpoints have their own instance or portion of those resources, locally. The distance that a resource request must travel is much shorter in a virtualized infrastructure.

Basically, the vaster and more varied an organization's edge "ecosystem" is, the less feasible it is to service that ecosystem from a few core locations. Disparate endpoints need to function more independently. When branches are farther from the root/trunk, they need to be able to hold more water and for longer.

Edge virtualization vs. core virtualization

Conceptually, edge virtualization contrasts with core data center virtualization in the following ways. Individual "instances" of edge virtualization:

• are on a much smaller scale;
• are located far from the edge's control point without local support;
• may have limited communications bandwidth, which may also suffer significant latency; and
• are required to support a wide range of workload types that interface with real-world peripherals using a wide range of technologies, many of which may be legacy or nonstandard.

Technically, the "virtualization" part of edge virtualization is not that different from traditional core virtualization in the data center.

The main difference comes from the unique compute challenges at the edge, which are often different than that of the data center. Edge virtualization requires integration with internet of things (IoT) systems, retail systems, specialized client devices and peripherals.

Edge virtualization integrates old and new peripherals regardless of heritage. Peripherals are separate devices (scanner, printer, etc.) that remain connected to network endpoints where they are needed. All their workloads are securely connected to the virtualized software running on a local server. The workload of each endpoint is moved onto a virtualized server.

Use cases

Any massively distributed business or organization may benefit from network virtualization at the edge. It can be used in a variety of industries, such as:

• Retail
• Hospitality
• Healthcare
• Banking, specifically retail banks that still maintain a branch network
• Infrastructure companies with many remote outstations like water, electricity and telecoms

For example, in a retail environment, a virtualized POS requires seamless integration to peripherals such as the barcode scanner, magnetic stripe reader, cash drawer or receipt printer. A virtualized software stack unifies these and allows end users to seamlessly integrate them regardless of any variation or incongruence in their underlying hardware.

In this example, POS software is hosted on a small virtualized server positioned in a business's back office, out of sight from customers. Essentially, the software is decoupled from the hardware. The user could theoretically remove the physical disk from the POS hardware (the most common point of failure) and run the POS as a virtual machine (VM) on the server.

How edge virtualization ties into 5G

5G requires a massively distributed infrastructure to function. The edge nodes in 5G will interface with a variety of peripherals, IoT and networked devices of various heritages. Because a 5G network needs to be highly distributed, each network node or edge device needs to have local virtualized resources. A 5G network would be impossible to implement if each node needed to access resources from a centralized location for updates or maintenance, because of the sheer number of edge nodes and their complex reliance on each other to function.