An Extensible Edge Computing Architecture: Definition, Requirements and Enablers

Abstract

Cloud computing is highly being used for several years for various purposes. From daily tasks, such as reading e-mails, watching videos to the factory automation and device control, it changed where the data is being processed and how it is accessed. However, increasing number of connected devices brings problems, such as low Quality of Service (QoS) due to infrastructure resources and high latency because of the bandwidth limitations. The current tendency to solve the problems that the Cloud computing has is performing the computations as close as possible to the device. This paradigm is called Edge Computing. There are several proposed architectures for the Edge Computing, but there is no an accepted standard by the community or the industry. Besides, there is not a common agreement on how the Edge Computing architecture physically looks like. In this paper, we describe the Edge Computing, explain how its architecture looks like, its requirements, and enablers. We also define the major features that one Edge Server should support.

Full text

An Extensible Edge Computing Architecture: Definition, Requirements and Enablers
Volkan Gezer and Jumyung Um and Martin Ruskowski
Innovative Factory Systems (IFS)
German Research Center for Artificial Intelligence (DFKI)
Kaiserslautern, Germany
Emails: {name.surname}@dfki.de
Abstract—Cloud computing is highly being used for several years
for various purposes. From daily tasks, such as reading e-mails,
watching videos to the factory automation and device control,
it changed where the data is being processed and how it is
accessed. However, increasing number of connected devices brings
problems, such as low Quality of Service (QoS) due to infras-
tructure resources and high latency because of the bandwidth
limitations. The current tendency to solve the problems that the
Cloud computing has is performing the computations as close as
possible to the device. This paradigm is called Edge Computing.
There are several proposed architectures for the Edge Computing,
but there is no an accepted standard by the community or the
industry. Besides, there is not a common agreement on how the
Edge Computing architecture physically looks like. In this paper,
we describe the Edge Computing, explain how its architecture
looks like, its requirements, and enablers. We also define the
major features that one Edge Server should support.
Keywords–Edge computing; requirements; enablers; Fog com-
puting.
I. INTRODUCTION
With the increased tendency towards Internet of Things
(IoT), number of connected devices to the Internet are in-
creasing day by day. In 1992, the connected devices count was
around one million which went up to 500 million in 2003 with
increased usage of notebooks. Later, IoT became even more
popular and made three billions of devices connected. In 2012,
with the inclusion of wearable devices this number went high
as 8.7 billion. In 2013, this number was 11.2 billion thanks
to connected home appliances and in 2014, 14.4 billion with
smart grids. The numbers increased in the upcoming years due
to involvement of small personal objects, such as toothbrushes,
traffic lights, and table watches. Finally, even door levers are
expected to be part of smart objects in 2020 [1].
Connected devices are expected to be around 50 billion
by 2020 [1][2]. This number is high as the Cyber-Physical
Systems (CPS) and more intelligent components being used
even for simple tasks. Using different standards, a single in-
frastructure to keep the system reliable is becoming even more
complex, causing difficult and costly maintenance. Relying on
a single information technology (IT) infrastructure can also
increase the downtime of communication which disrupts the
service leading to non-productive time. The bandwidth for
communication is also becoming a problem to transmit that
amount of data.
Cloud Computing [3] is an emerging technology which
allows machines/people to access the data ubiquitously. It
enables on-demand sharing of available computing and storage
resource among its users which could be either human or
machine, or even both. Today, it is even possible for a simple
device to share its status or get information over Internet with
millions of users. In Cloud Computing, the communication
between the device and the infrastructure which provides the
service is direct, without involvement of other tiers. However,
increased usage of Cloud increases latency and the load on
the server and on the network. Having billions of devices and
processing the data produced by each of them is a troublesome
task for centralized systems [4]. Figure 1 shows some examples
for Cloud Computing, such as E-Mail services, Cloud Storage
systems, Video hosting web sites, etc.
A layer is a logical organisation of set of services, devices,
or software with the same/similar specific functionality, mainly
defined for abstraction of tasks. A tier is, however, a physical
deployment of layers for scalability, security and to balance
performance [5].
Figure 1. Some of daily usage examples of Cloud Computing, such as
e-mails, music/video streaming, and data storage.
Edge Computing is a recent paradigm, which moves com-
puting application and services from centralized units into the
logical extremes or at the closest locations to the source and
provides data processing power there. It adds an additional tier
between the Cloud and the end-devices as depicted in Figure
2. Increase in Edge nodes within a location will reduce the
number of devices connected to a single Cloud and eliminate
the problems of the Cloud Computing. Examples to Edge
Computing can be listed as Smart Cities, Machine to Machine
communication, Security Systems, Augmented Reality, Wear-
able Health Care Systems, Connected Cars, and Intelligent
Transportation. For example, a plane produces gigabytes of
148Copyright (c) IARIA, 2017. ISBN: 978-1-61208-598-2
UBICOMM 2017 : The Eleventh International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies

data per second [6], which cannot be handled by a single base
infrastructure due to bandwidth limitations. Another example
is a Formula One car which produces approximately 1.2 GB/s
data [7] that requires gathering, analysis, and acting in-time to
stay competitive in the race [8]. Edge Computing is believed to
solve these issues by aggregating and pre-processing the data
in Edge, before transmitting to the Cloud or even deciding the
next steps on the Edge.
Figure 2. A simplified version of communication using Edge Computing.
Both Edge Computing and Cloud Computing are part of
Internet of Things (IoT) and allow accessibility of the data
ubiquitously. To build an architecture, the issues on the current
Cloud or IoT systems must be identified, requirements must
be specified, enabling technologies must be listed, and then a
concept must be given. Later, the concept can be implemented
in an architecture, validated, and evaluated.
This paper presents an ongoing work on Edge Computing
with its clear description. It also explains its requirements and
enablers to solve the introduced issues because of high usage
of Cloud and IoT.
The paper is structured as follows. In Section II, a short
overview on related work in Edge Computing domain is given.
In Section III, the concept of Edge Computing is explained.
Later, in Section IV, its requirements, and in Section V, en-
ablers are explained. In Section VI, the major functionalities of
the proposed architecture is explained. The paper is concluded
in Section VII with the future work.
II. RE LATE D WOR K
Although usage of the term “Edge Computing” is recent,
there are already several proposed architectures available, each
considering different aspects to meet the requirements of
the Edge Computing. Below, some of the existing proposed
architectures will be discussed.
The architecture proposed by IBM considers the require-
ments for autonomy and self-sufficiency of production sites.
The architecture is three-layered to balance the workload
between the Edge, the Plant, and the Enterprise. The challenges
of the architecture are listed as productivity gains for high
throughput, failure prevention for reliable system and high
product quality, and flexibility while hiding the complexity
and allowing reconfiguration without a lot of effort [9].
Another reference architecture is proposed by OpenFog
Consortium [10]. This architecture names the core principles
as pillars. Pillars group requirements within their scope. These
pillars are Security, Scalability, Openness, Autonomy, Agility,
and Programmability. OpenFog Reference Architecture is pro-
posed by covering industrial use cases.
Another recent initiative to build a common platform for
Industrial IoT Edge Computing is EdgeX Foundry [11]. It was
launched by Linux Foundation and initial contribution made
by Dell. However, similar to OpenFog Consortium, it is also
open for new memberships. EdgeX Foundry is a vendor-neutral
open source software platform that interacts at the Edge of the
network. It defines its requirements in architectural tenets as
follows: platform agnostic in terms of hardware and operating
system, flexible in terms of replacability, augmentability, or
scalability up and down, capable in storing or forwarding data,
intelligent to deal with latency, bandwidth, and storage issues,
secure, and easily manageable. A similar framework called
Liota is being developed by VMware and it also aims at easy
to use, install, and modify. Secondarily, it targets for a general,
modular and enterprise-level quality. This framework is also
open source and governed by VMware [12].
The aim in this research is not simply to build another
architecture, but to analyse the existing architectures and
consider industrial requirements to make up a generic reference
architecture which is vendor-independent and extensible. The
architecture is also able to execute real-time tasks. To the
best of our knowledge, this is not considered in any of the
aforementioned reference architectures.
III. CON CE PT
One of the main goals of Edge Computing is to reduce
latency and to keep the Quality of Service (QoS) as high
as possible. As seen in Figure 1, in Cloud Computing, the
Cloud infrastructure communicates with the end-devices di-
rectly. Edge Computing intends to solve the issues of Cloud
Computing or IoT by adding an additional tier between the
IoT devices and back-end infrastructure for computing and
communication purposes. As depicted in Figure 3, this tier also
has intermediate components for the first gathering, analysis,
computation of the data. These intermediate components are
called Edge Servers. Several architecture types for IoT-enabled
applications are proposed [13]. In this paper, a three-tier
architecture is used.
Figure 3. Edge Computing is an additional tier between Cloud and the
Devices. The Edge Servers can be in the same or different physical locations.
As seen in Figure 3, the proposed architecture for Edge
149Copyright (c) IARIA, 2017. ISBN: 978-1-61208-598-2
UBICOMM 2017 : The Eleventh International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies

Computing consists of Cloud Tier,Edge Tier, and Device
Tier. In the Device Tier, there are end-user devices. The green
blocks in the Edge Tier are Edge Servers. These servers gather,
aggregate, analyse, and process the data before offloading
them to the Cloud Tier. The end-devices can be in the same
location, or in different physical locations as depicted in the
figure. When an end-device needs to communicate with the
Cloud, first, the request is sent to the Edge Server which is
at the closest location. Then, if the Edge Server is capable of
completing the task by itself, it automatically handles the data
and responds to the end-device with the result. If not, the data
is offloaded to another server in the same tier provided that
it exists. Otherwise, the data is offloaded to the Cloud. The
decision process is made by considering available resources in
other available servers in the same network, physical distance,
and time requirements.
Assume that the Cloud provides functionalities A and B.
When one of the devices in Physical Location X intends to
do task B, first the data is passed to the server #1. Since this
server is not capable of performing this task, it passes the data
either to the Cloud. As Cloud is capable of performing task B,
the data is processed here and sent back to the originating end-
device. The challenge here is to decide on functionalities in
the Edge Tier by keeping the costs at minimum and the QoS
at maximum. However, deciding on the count and available
resources of Edge Servers are also big challenges and big
trade-offs. There are several aspects to consider before passing
the data to the Cloud. For example, if a device located at Y
needs task A to be done, and if the Edge Server #2 is busy
with servicing other two connected devices, another trade-off
will be existent. In this case, the server #2 needs to offload the
task either onto server #1 or the Cloud. However, depending
on the urgency of the task, the server #2 needs to calculate a
function to decide on the best recipient of the data. According
to this, the function should consider the priority of the task,
resource utilization of the servers, computing cost for the task,
and the physical distance or distance cost of the servers that
is going to be used.
IV. REQUIREMENTS
Edge Computing is a paradigm which uses Cloud Comput-
ing technologies and gives more responsibilities to the Edge
tier. These responsibilities are namely, computing offload,
data caching/storage, data processing, service distribution, IoT
management, security, and privacy protection [4].
Without limiting the Cloud Computing features, Edge
Computing needs to have the following requirements, some
of which are also defined for Cloud Computing [14][15]:
1) Interoperability: Servers in Edge Computing can con-
nect with various devices and other servers. In Cloud Comput-
ing, IoT allows countless number of devices to communicate
with humans or each other. This creates a big market for
manufacturers of these devices. For this reason, there is the
issue of interoperability with connected device using different
communication protocols. Advanced Message Queuing Proto-
col (AMQP), Message Queue Telemetry Transport (MQTT),
and TCP/IP are widely used and should supported by Edge
Computing. Using a widely-used and widely-known standard
will remove the technology and language barriers, increasing
interoperability among the devices.
2) Scalability: Similar to Cloud services, Edge Computing
will also need to be adapted for the size of its users and
sensors. First deployment enables small number of users and
devices while few Edge Servers should handle higher number.
Additional deployment of Edge Servers is costly and small
number of Edge Servers is desirable in terms of economical
aspects. For this reason, high scalability is also mandatory.
3) Extensibility: Computing technology is developing
rapidly. After 2-3 years of deployment, clock speeds, memory
size and program size increase, too. Easy deployment of new
services and new devices with small effort is required for
essential goal of Edge Computing. New functions and devices
should be integrated without (re)configuration of the Edge
network. Therefore, the system should allow extensibility with
hardware and software components.
4) Abstraction: For the seamless control and communica-
tion, the abstraction of each Edge Node and group of nodes is
required. Moreover, abstraction helps the topology of an Edge
network to be flexible and reconfigurable. Fundamentally, an
Edge node is located between device tier and Cloud tier. In
other words, an Edge tier is a border between Information
Technology (IT) and Operational Technology (OT). This tier
can consist of one or more Edge nodes and groups. In this
case, one Edge node of the group can share tasks or nodes
in the group can be prioritized. Utilization of Application
Programming Interfaces (APIs) in abstraction is useful to
provide backward compatibility for the new functionalities or
big changes in the architecture.
5) Time sensitiveness: Below OT, the operations may be
near-real-time or real-time. Edge Computing is expected to
solve time issues which Cloud computing cannot guarantee.
Unlike Cloud Computing, physically close distance is one
strength of reliable and fast communication without worry-
ing about traffic problem. Video streaming service is one
of expected applications of Edge Computing. It is required
for real-timeness of the service provision. In addition, time-
sensitiveness adds big benefits to providers of reactive services,
such as location-based advertisements and user-status based
guide systems.
6) Security & Privacy: Using Cloud Computing services
has a trade-off for enterprises like manufacturing and high-
tech companies because there is a concern about the leakage
of high knowledge and business activities outside their own
organization. Edge Computing is a way to secure data contents,
which is different from firewall which only controls external
access into the network. It is also important to isolate the data
by preventing access from even non-authorized users.
7) Reliability: Edge Servers provide real-time or non-real-
time control for the devices. Real-time tasks may be vital
which involve human safety. Therefore, it is vital to have
a reliable system which reacts when it is needed and how
it is needed. The physical reliability requirements for Edge
servers providing services is similar to Cloud Computing.
Harsh environments, such as factories and construction yards,
require water-proof ceiling, fanless computers and dust-proof
system. In power plant, magnetic shield is equipped by sensor
gateways.
8) Intelligence: Multi-sensor generates tremendous amount
of data and uploads into Cloud, directly. It causes network con-
gestion and heavy load on the Cloud server. Edge Computing
150Copyright (c) IARIA, 2017. ISBN: 978-1-61208-598-2
UBICOMM 2017 : The Eleventh International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies

supports first and second filtering of these data by converting
into higher level of data contents. Data filtering is implemented
by rule-based engines or machine learning algorithms. In the
case of multi-camera system like security systems, Edge Com-
puting supports image processing, computer vision and enables
object detection before transferring the data into the Cloud.
Another example is predicting the failure or abnormalities in
a production line by analysing the sensor data and taking the
precautions for prevention or informing the user. These kinds
of intelligent functions are necessary for Edge Computing.
9) Power: Unexpected shutdown or blackout is the cause
of breakdown of Edge Server. Uninterruptible power supply
(UPS) is required to give an ample amount of time to protect
the electronic units and data storage in case of an unexpected
shutdown due to power outage.
V. E NA BL ER S
Edge Computing uses wide range of technologies and
brings them together. Within this domain, Edge Computing
utilizes many technologies, such as wireless sensor networks
(WSN), mobile data acquisition, mobile signature analy-
sis, Fog/Grid Computing, distributed data operations, remote
Cloud services, etc. Additionally, it combines the following
protocols and terms:
1) 5G communication: It is the fifth generation wireless
system which aims at higher capacity, lower power consump-
tion, and lower latency compared to the previous generations.
Due to increased amount of data between the data, 5G is
expected to solve traffic issues which arose with the increased
number of connected devices.
2) PLC protocols: Object Linking and Embedding for
Process Control Unified Architecture (OPC-UA) is a protocol
developed for industrial automation. Due to its openness and
robustness, it is widely used by industries in the area of oil
and gas, pharmaceutical, robotics, and manufacturing.
3) Message queue broker: MQTT and TCP/IP are pop-
ular message protocols of smart sensors and IoT devices.
Supporting these message brokers, Edge Computing increases
the device count that it connects. For the problem of MQTT
security, AMQP is useful in the communication with Cloud
Computing server.
4) Event processor: After messages of IoT arrive in the
Edge server, event processor analyses those messages and
creates semantic events using pre-defined rules. EsperNet,
Apache Spark, and Flink are some examples for this enabler.
5) Virtualisation: Cloud services are deployed as virtual
machines on a Cloud server or clusters. Using virtual machines
allow running multiple instances of operating systems (OS) on
the same server.
6) Hypervisor: As well as virtual machine, performance
evaluation and data handling are required and realized by
hypervisor to control virtual machines in the host computer.
7) OpenStack: Managing multiple resources could be chal-
lenging. OpenStack is a Cloud operating system that helps
control of pools of computing and storage resources at ease
through a control panel and monitoring tools.
8) AI platform: Rule-based engine and Machine learning
platform supports data analysis in local level. As stated in
Section IV, this is quite important to reach one of the goals
of Edge Computing which is to gather, analyse, and perform
the first filtering of the data.
9) Hyperledger: Blockchain technology is currently used
for highly sensitive areas, such as digital currencies like
BitCoin. It is also considered as useful for the data protection
in Cloud Computing. By using this technology, secure data can
be shared with external persons and servers with high security.
10)Docker: Virtual machines work with installation of
operating systems. Unlike virtual machines, Docker is a Con-
tainer as a Service (CaaS) which can use a single shared
operating system and run software in isolated environment.
It only requires the libraries of the software which makes it a
lightweight system without worrying about where the software
is deployed.
VI. ARCHITECTURE DESIGN
Edge Computing adds an additional tier between the Cloud
and IoT devices for computing and communication. The data
produced by the devices themselves are not directly sent to
the Cloud or back-end infrastructure, but initial computing is
performed on this tier. Considering the number of connected
devices and the data they produced, this tier is used to
aggregate, analyse, and process the data before sending it into
the upper layer, the infrastructure.
Figure 4 depicts the proposed core functionalities for an
Edge Server.
Figure 4. View of the proposed extensible Edge server architecture with its
major functionalities, where green blocks extend the functionalities for the
blue core node.
The proposed Edge Server architecture is to be designed
modular and should provide functionalities for real-time and
non-real-time control, as well as real-time communication.
Core node runs on an operating system and tracks resources
and makes decisions on where to execute a task. In the
proposed architecture, addition of a new hardware or software
modules enable new functionalities and improve the usability
of the server. For example, in the case that machine learning
algorithms are desired to be executed on the server, connecting
a dedicated artificial intelligence (AI) module with dedicated
Graphics Processing Unit (GPU) should require none to min-
imal configuration to be active.
151Copyright (c) IARIA, 2017. ISBN: 978-1-61208-598-2
UBICOMM 2017 : The Eleventh International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies

As mentioned in Section IV, scalability is quite important
to accomplish the tasks. In the scope of scalability, one server
is expected to be aware of its neighbouring servers along with
their functionalities. Using the previous example, in case an AI
module is connected to one server, other servers are informed
with this functionality and they can utilize this server more
often for AI-related tasks. The decision, of course, depends on
the conditions required by the task, such as deadline.
VII. CONCLUSION AND FUTURE WOR K
Edge Computing is a recent term which moves the services
from the Cloud to the device as close as possible. It is a
borderline between the Cloud and the device tier. Although
the Cloud Computing has brought many advantages in the
previous years, increased number in the connected devices
raised some issues, such as latency and low QoS problems.
Edge Computing is believed to solve these issues by analysing
the issues and considering the requirements of real world use
cases.
This paper showed an ongoing work on how Edge Com-
puting physically looks like together with its requirements and
enablers. It also explained the basics on how the communica-
tion between the end-devices and Edge servers are expected
to be.
There are already several existing proposed architectures in
the domain of Edge Computing, such as EdgeX Foundry, Liota,
and OpenFog Reference Architecture. Although they are also
extensible and they allow inter-connectivity, they do not talk
about the real-timeliness of the architectures. This work will be
focusing on real-time computing and communication for the
given tasks. Of course, it will also be available for non-real-
time tasks. The work is being developed by considering the
real-world use cases of the industrial partners. The validation
will be performed with these use cases and the comparison
with the legacy systems will be made.
In the future, internal software and hardware components
for the Edge Server will be decided. Later, they will be
simulated as an initial work for the architecture design. Next,
the software components will be individually implemented in
the simulation environment. By analysing the simulator results,
a hardware benchmarking will be performed and a hardware
will be chosen to be used as the Edge Server solution. The
final task will be to realize the components by deploying them
on the chosen hardware.
ACKNOWLEDGMENT
This research was funded in part by the H2020 program
of European Union, project number (project FAR-EDGE).
The responsibility for this publication lies with the authors.
The project details can be found under project website at:
http://www.far-edge.eu
REFERENCES
[1] NCTA, “The Growth of The Internet of Things,” Infographic,
May 2014, [retrieved: Sep 2017]. [Online]. Available: https://
www.ncta.com/platform/industry- news/infographic- the-growth- of- the-
internet-of- things/
[2] D. Evans, “The Internet of Things - Cisco,” Cisco, White Paper, April
2011, [retrieved: Sep 2017]. [Online]. Available: https://www.cisco.
com/c/dam/en us/about/ac79/docs/innov/IoT IBSG 0411FINAL.pdf
[3] M. Peter and G. Timothy, “The nist definition of cloud computing,”
in National Institute of Standards and Technology Technical report,
September 2011.
[4] W. Shi, J. Cao, Q. Zhang, Y. Li, and L. Xu, “Edge Computing: Vision
and Challenges,” IEEE Internet of Things Journal, vol. 3, no. 5, October
2016, pp. 637–646.
[5] R. Lhotka, “Should all apps be n-tier?” Blog, 2005, [retrieved:
Sep 2017]. [Online]. Available: http :// www.lhotka .net /weblog /
ShouldAllAppsBeNtier.aspx
[6] S. Higginbotham, “Sensor Networks Top Social Networks for Big
Data,” Article, 2010, [retrieved: Sep 2017]. [Online]. Available:
https:// gigaom.com/2010/09/13/sensor- networks- top-social-networks-
for-big-data-2/
[7] T. Valich, “Big Data In Planes: New P&W Gtf Engine Telemetry
To Generate 10GB/s,” Article, 2015, [retrieved: Sep 2017]. [Online].
Available: https://vrworld.com/2015/05/08/big-data-in- planes-new- pw-
gtf-engine- telemetry-to- generate-10gbs/
[8] F. Bi, “How Formula One Teams Are Using Big Data To
Get The Inside Edge,” Article, 2017, [retrieved: Sep 2017].
[Online]. Available: https://www.forbes.com/sites/frankbi/2014/11/13/
how-formula-one-teams-are-using-big-data-to-get-the-inside-edge/
[9] I. C. A. Center, “IBM: Internet of Things,” Cloud Garage Method,
2017, [retrieved: Sep 2017]. [Online]. Available: https://www.ibm.com/
devops/method/content/architecture/iotArchitecture/industrie 40
[10] “OpenFog Consortium Reference Architecture,” Website, 2017,
[retrieved: Sep 2017]. [Online]. Available: https : / / www.
openfogconsortium.org/ra/
[11] “EdgeX Foundry Architectural Tenets,” EdgeX Foundry Wiki, 2017,
[retrieved: Sep 2017]. [Online]. Available: https://wiki.edgexfoundry.
org/display/FA/Introduction+to+EdgeX+Foundry
[12] “VMware Introduces Liota,” Website, 2017, [retrieved: Sep
2017]. [Online]. Available: https://www.vmware.com/radius/vmware-
introduces-liota- iot-developers-dream/
[13] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and
M. Ayyash, “Internet of things: A survey on enabling technologies,
protocols, and applications,” IEEE Communications Surveys Tutorials,
vol. 17, no. 4, 2015, pp. 2347–2376.
[14] G. Orsini, D. Bade, and W. Lamersdorf, “Context-Aware Computation
Offloading for Mobile Cloud Computing: Requirements Analysis, Sur-
vey and Design Guideline,” Procedia Computer Science, vol. 56(1),
December 2015, pp. 10–17.
[15] J. Shamsi, M. A. Khojaye, and M. A. Qasmi, “Data-intensive cloud
computing: Requirements, expectations, challenges, and solutions,”
Journal of Grid Computing, vol. 11, no. 2, Jun 2013, pp. 281–310.
152Copyright (c) IARIA, 2017. ISBN: 978-1-61208-598-2
UBICOMM 2017 : The Eleventh International Conference on Mobile Ubiquitous Computing, Systems, Services and Technologies