Page 67 - Kaleidoscope Academic Conference Proceedings 2022

Extended reality – How to boost quality of experience and interoperability

The edge computing concept was initially envisioned as part network, can complicate the interaction between
of 5G deployments [9], [10] for drastically reducing the administrators and the infrastructure. Hence, there is a need
latency of cloud-based applications. But it did not prove to for a unified platform, e.g. in the form of a network operating
be the case as there are still major obstacles to overcome system, through which the interaction can be simplified in
before it becomes a reality. There needs to be seamless order to flexibly program the operation of resources and
integration between communication and computation services.
technologies, which have been traditionally evolving on their
own parallel paths. Carefully crafted management 5. NETWORK ADDRESSING
functionality can provide the right glue and can allow the two
to be operated and optimized in a unified manner. While the Internet Protocol (IP) has been adopted as the de
facto layer-3 protocol for Internet-based communications, it
To meet the performance requirements of future demanding has inherited several limitations of its original design as
services, telecommunication network and service scalability and applicability issues were not properly
management practices need to become more dynamic and considered at the time. Although IPv6 solves the problem of
flexible in nature and match the capabilities of those in cloud address depletion, it still inherits a number of the IPv4
domain. Going beyond the traditional telecommunication problems concerning, for example, the fixed number of
virtualization technologies and platforms, a cloud-native address bits and the fixed semantics associated with the
approach is needed, which hides the underlying addresses. These limitations have been demonstrated by the
heterogeneous infrastructure by operating on smaller emergence of several networking scenarios, which require
containers, such as those provided by Docker. This will significantly more flexibility in network addressing, both in
provide a level of abstraction above the compute terms of length and semantics.
infrastructure allowing applications to be formed from one
or more components executing within lightweight containers. 5.1 Low power IoT networks
The management logic for coordinating the container-
hosting locations, the allocation of user requests among the The low device complexity of IoT devices defines the
distributed set of application execution points, and the properties of communication technologies developed for this
decisions concerning the usage of resources will need to domain, which use short (e.g. 16-bit) addresses in order to
follow a multidimensional optimization framework. reduce the header size, communication overheads and
memory requirements. IPv4 32-bit addresses are already
4. NETWORK SOFTWARIZATION long, resulting in expensive operations in multi-hop routing
scenarios, while IPv6 128 bit-addresses worsen the problem
Although hardware implementations in the network are in and would not even fit within the maximum transmission
general very fast and have predictable behavior due to unit of some IoT protocols [19]. Header compression
exhaustive testing, they are rigid in terms of location and techniques can potentially decrease the IPv6 overhead and
available resources, and their update/maintenance tends to be fragmentation of packets would allow coping with the bulky
complicated. To overcome these limitations, the research protocol. These, however, involve power-hungry operations
community has been investigating software-based solutions inappropriate for resource-constrained IoT devices and
which can achieve greater agility and cost effectiveness. hence require the use of gateways.

The two main technologies that drive the so-called Due to its inherent rigidity in supporting only the single
"softwarization" of telecommunication networks are addressing semantic of topological location, IP protocol is
Network Functions Virtualization (NFV) [7] and SDN [12]. not able to meet the requirements of a wide range of IoT
The former allows us to replace network equipment, such as application scenarios in which addresses can take various
load balancers and firewalls, with software that executes on forms to identify, for example, communication endpoints,
commodity servers, while the control data plane separation physical objects, data types, and locations within a
in SDN moves control functions outside network devices and geographic area [1]. In addition, support for multiple
into dedicated controller entities. SDN largely simplifies semantics would be lost in a scenario where address
network management tasks and allows advanced network translation is needed when connecting multiple IoT networks
intelligence to be flexibly added to controllers without over the Internet. Flexibility in the address length, which can
needing to upgrade network devices. NFV, on the other hand, also be seen as elasticity in the address space, caters to an
allows network functions to be created/migrated according increasing number of specialized network deployments and
to the needs and for updates to be performed remotely, while would allow IoT networks to use the length that best fits their
the use of virtualization technologies enables the dynamic scale and constraints. In addition, flexible addressing can
scaling of resources allocated to network functions. enable seamless communication between nodes, bypassing
the need for expensive address mappings.
The significant advantages such technologies offer have
been driving the trend towards software-based solutions. The 5.2 Highly dynamic network topologies
fact, however, that multiple technologies are needed to
manage physical and virtual resources, alongside the The semantically rigid nature of IP addresses, which refers
imminent integration of computational capability in the only to the topological location of a network interface, poses

– 21 –