Collaborative Computing Paradigms: A Software Systems Architecture

for Dynamic IoT Environments

Prashant G. Joshi and Bharat M. Deshpande

Department of Computer Science & Information Systems, BITS Pilani K K Birla Goa Campus, Goa, India

Keywords:

Collaborative Computing Paradigms, IoT, Computing Paradigms, Cloud Computing, Edge Computing,

Mist Computing, Software Architecture, Architecture Models, System Software Architecture, Collaborative

Computing, Collaborating Paradigms.

Abstract:

Connected systems are omnipresent, are used to monitor and control remotely, collect data and information.

A variety of software systems architectures are designed which exploit computing paradigms – Edge, Fog,

Mobile and Cloud – that process and analyse the data. Such an analysis is pivotal in decision making to in-

crease operational efﬁciency. IoT has transformed industries like logistics, healthcare, industrial automation

and agriculture and continues to reﬁne decision making process ultimately to enhance the systems operations

and efﬁciency. Expanding service capability of systems, has been topic of academic research and, rests on the

foundation of bringing all resources in uniﬁed resource pool and make different computing facilities collabo-

rate. Making the different computing facilities to collaborate to realise this has been part of many theoretical

and experimental studies. Industry applications have adopted such systems architectures that has enhanced the

applications capabilities. This paper proposes a collaborative and uniﬁed method of system software architec-

tures, for IoT environments, that leverage collaboration among computing paradigms. With a view to expand

services, as and when needed, a uniﬁed, dynamic and distributed analytics software systems architecture was

explored and experimented. Proposed collaborative method is validated through its application for vehicle and

driver behaviour and data center cooling systems.

1 INTRODUCTION

Ubiquitous deployment of smart and connected de-

vices has led to an exponential increase in the need

for processing the collected data. The trend of ap-

plications based on collected data has been on a sig-

niﬁcant rise and is expected to continue growing at

a substantial rate. The collected heterogeneous and

high-volume data, gathered at frequent intervals, is

subsequently subjected to extensive processing, en-

abling the derivation of variety of analytics crucial for

decision-making (Donno and et al., 2019). This per-

vasive deployment has resulted in various successful

implementations, particularly evident in areas such as

vehicle tracking, where in-depth analysis of vehicle

and driver behavior has been accomplished (Malekian

and et al., 2015) (T

urker and et al, 2016) . All such

systems are built around a generic software systems

architectures where the smart devices sense and en-

able data collection which is then processed by trans-

mitting upstream to Cloud servers, for computing,

over various networking interfaces.

In most real-world applications, it is necessary for

collected data to be processed faster and results of

the processing, which may be alerts or notiﬁcations,

transmitted quickly. Such applications ﬁnd the typi-

cal architectures in which device transmitting data to

cloud for processing and generating alerts to be cum-

bersome. Edge computing paradigm is a paradigm of

choice for many such real-world applications. With

introduction of edge computing, typically in an hier-

archical or layered approach along with the cloud, the

computing needs for use cases are typically ﬁxed at

design time, which separates the processing into what

is computed at the edge and what is done on the cloud.

As discussed in (Kaushik and Naik, 2023), the tem-

perature is sensed and transmitted to the cloud server

where the data is processed and the necessary param-

eters are communicated back to the edge to control

the temperature by switching the air conditioners.

For most applications, Cloud computing (Peter

and Timothy, 2011), which provides a cost-effective,

scalable and always available infrastructure for com-

puting, dependable and managed software, elastic

Joshi, P. and Deshpande, B.

Collaborative Computing Paradigms: A Software Systems Architecture for Dynamic IoT Environments.

DOI: 10.5220/0012473000003645

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 12th International Conference on Model-Based Software and Systems Engineering (MODELSWARD 2024), pages 297-306

ISBN: 978-989-758-682-8; ISSN: 2184-4348

297

storage, and resource virtualization, is used. Over

time the demand for computing has grown signiﬁ-

cantly with the increase in the volume of heteroge-

neous data collected at high frequencies from variety

of devices, and real-time responses, thus giving rise

to newer paradigms - Edge, Mist, Fog and Mobile -

which have been used to overcame some of the limi-

tations of Cloud computing (Iorga and et al., 2018).

Every paradigm that emerged focuses on data pro-

cessing closer to the source or data or the user, thus

addressing the challenge of near real-time response

times, interactive and on-line applications.

Software systems architectures of today, as dis-

cussed in (Garc

es and et al, 2021), involve numerous

interconnected components such as hardware, soft-

ware, networks and users of the system. Past years

have seen a very signiﬁcant advancement in commu-

nication and computing technologies, that have led to

a transformative impact on computing, storage, wired

and wireless networks, and application deployment.

Such an impact is observed across domains like con-

sumer, commercial and industrial.

With the increasing complexity of applications,

processing in real-time and faster decision making,

the standardization of architectures has become a ne-

cessity. Both academia and industry have developed

their own reference architectures to address this need.

Academia has primarily focused on characteristics

of reuse and generating domain speciﬁc knowledge

(Garc

es and et al, 2021). On the other hand, the in-

dustry has created reference architectures to deliver

systems and solutions tailored to speciﬁc applications

with a certain level of maturity. Over time, the in-

dustry has developed reference architectures for vari-

ous applications and gathered best practices and refer-

ences for designing new systems in different domains.

Irrespective of whether the architectures origi-

nated in academia or industry, all software system

architectures have leveraged and exploited the avail-

able computing paradigms, and industry driven appli-

cations and real-world applications have been devel-

oped using them.

Paper is organised as follows. Section II pro-

vides a brief on the Computing Paradigms Land-

scapes which compares the current paradigms. Sec-

tion III identiﬁes the Challenges and Issues in the IoT

environments summarising them based on the pub-

lished research, Section IV proposes a Collaborative

Computing Architecture to overcome the challenges

and issues with a focus on dynamic IoT environments.

Section V describes and discusses the applications

and use-cases highlighting the characteristics of the

proposed Collaborative Computing Paradigm. Sec-

tion VI provides a conclusion and Section VII closes

with the future work.

2 COMPUTING PARADIGMS

LANDSCAPE

Cloud, Edge, Fog, Mist and Mobile computing have

been proposed, which are characterised by variety of

interfaces and access technologies, and varying ca-

pacity of resources like processing power, storage, in-

terface and closeness to the source.

In the TABLE 1, a summary of compari-

son of these characteristics for various computing

paradigms, considered for this research, has been

shown.

Additionally, the TABLE 2 summarises the

strengths and weaknesses of each computing

paradigm, which is useful and typically considered

when choosing one or more computing paradigms for

a particular application or a use-case.

2.1 Computing Paradigms

Collaboration - Challenges & Issues

Computing paradigms differ signiﬁcantly based on

infrastructure, storage, computing capability, ac-

cess technology, resource management and service

provider. It is clear from the literature as well as from

the origin of the computing paradigms that different

paradigms are designed for different services. For all

of the computing paradigms, theoretical studies often

ignore their differences and focus on their strengths

put to use in realising an application.

Collaboration among the paradigms will require

resources of one paradigm to be shared with other

paradigm, which is not easy and it will then be dif-

ﬁcult to converge these resources in to a uniﬁed pool.

Studies show mechanism of virtualisation of each of

the paradigms to such a level that one paradigm can

collaborate with the other and can be an effective

way of collaboration among paradigms. Such a tech-

nique is primarily designed to shield the heterogene-

ity among the computing paradigms. (Cai and et al,

2023; Nascimento and et al, 2023; Lewandowski and

et al, 2020)

In this paper, we propose to leverage the function-

ality provided by each of the paradigms, use the het-

erogeneity with a view to enable the high priority use-

cases of the application.

MODELSWARD 2024 - 12th International Conference on Model-Based Software and Systems Engineering

298

Table 1: Comparison of Computing Paradigms.

Computing

Paradigms

Salient Features

Cloud (CC) Provides nearly unlimited and on-demand capacity of computing and storage, thus ser-

vices are scalable and expandable. All resources are available over standard network

interfaces (Peter and Timothy, 2011).

Fog (FC) Provides computing, storage and network connectivity services in close proximity to

data sources and users. Thus, has a lower latency and enhances reliability in compari-

son to cloud computing (Iorga and et al., 2018).

Mist (MiC) Provides lightweight processing and limited storage closer to the sensors and smart

devices. Mist computing devices are based on microprocessors or microcontrollers

and are developed to augment processing capabilities reducing the load on Fog and

Cloud computing (Escobar and et al, 2022).

Edge (EC) Provides functionalities of the Cloud to the edge in form of mini clouds which have

the capability of transfer of data, it’s processing and storage at the edge of the network.

Analysis is done in real-time without latency and facilitates quicker data processing

and content delivery (Fern

andez and et al, 2018).

Mobile (MC) Portable devices with limited computing capability such as Mobile phones, tablets,

and wearables. MC nodes can move across the environments and can feature multiple

connectivity interfaces (Mahmoudi and et al., 2018).

Table 2: Computing Paradigms - Strengths and Weaknesses.

Paradigm Real-Time Processing Power Storage Capacity Scalability Interfaces

Cloud Low High High High Low

Fog High Moderate Moderate Moderate Moderate

Mist High Moderate Moderate Moderate High

Edge High High Low High Moderate

Mobile Moderate Low Low Low Highest

3 INTERNET OF THINGS (IoT)

ENVIRONMENT -

CHALLENGES AND ISSUES

IoT environments are highly demanding and require

a dynamic model of software system architecture. In

applications like autonomous vehicles, smart cities,

agriculture, healthcare, large amounts of heteroge-

neous data from variety of sensors is collected. This

data needs to be effectively processed to fulﬁl the de-

mands of the application and stored for future use.

All IoT environments require end to end connectiv-

ity with the devices and all participating components

to provision, upgrade and manage the components.

Applications today still are dominated by an IoT

device (IoTD) and cloud computing paradigm. De-

vices use network interfaces to transmit data to the

cloud, where the data is stored and analysed. With the

need for overcoming the challenges of real-time, het-

erogeneity and low latency proposed in (Bonomi and

et al, 2012) Fog Computing, a high virtualised plat-

form which is spatially close to the devices and pro-

vides compute, storage and networking services be-

tween the devices at the edge and the cloud computing

core. The paper discusses fog computing and the IoT

in areas of connected vehicle, smart grid, and wire-

less sensor and actuator networks. This paved way to

build a substantial body of work with fog computing

at its’ core. An interplay between fog and cloud is

described, but not developed any further in the paper.

Fog computing has been classiﬁed and challenges

summarised in (Carla and et al., 2019) and architec-

tural challenges to make the system ultra-responsive

have been identiﬁed. Authors have indeed done a

wide and deep survey of the research work and have

classiﬁed the same in major areas Architecture and

Algorithm. A set of evaluation criteria like hetero-

geneity, QoS management, scalability, mobility, fed-

eration and interoperability has been summarised and

all the classiﬁed architectures and algorithms have

been evaluated based on these criteria. Architec-

ture based classiﬁcation has been further broken into

end-user application agnostic and application speciﬁc

while the algorithms have been further classiﬁed into

computing, content storage and distribution, impact

Collaborative Computing Paradigms: A Software Systems Architecture for Dynamic IoT Environments

299

in energy consumption, and speciﬁc end-user algo-

rithms. Authors observe that there needs to be fo-

cused and deep research in the area of federation of

Fog/Edge and Cloud. Furthermore, a complete inter-

play of IoT, Fog, cloud and Mobile computing has not

yet been attempted.

In the paper (Vasconcelos. and et al., 2019)

an algorithm is proposed to make a choice between

cloud, fog or mist computing based on cost, band-

width and latency criteria. Simulation results sug-

gest that the algorithm choice is adequate. However,

this is still a choice to be made at design time. This

area has been researched with proposals for workload

engineering, using OpenStack-based middleware, de-

scribed in (Giovanni and et al., 2019) along with the

advantages of workload engineering. As the pro-

posed solution is using software at various levels,

the processing exhibits software-deﬁned characteris-

tics which deﬁnitely show the promise of such tech-

niques. Viability of the solution is evaluated based on

the case study on an intelligent surveillance system,

but does not detail out the other aspects of orches-

tration policies, security, and strategies which are left

for future work. A uniﬁed model for Mobile-Edge-

Cloud continuum which uses Function-as-a-Service

in order to bring the computation in form of micro

services and an adaptive resource allocation for com-

puting ofﬂoading has been discussed in (Baresi and

et al., 2019), where edge servers are modelled as a set

of linear systems.

A collaboration between cloud, fog and IoT and

its effects on performance has been detailed in (Mo-

hammad and et al, 2018). Authors identify the inte-

gration issues within Cloud-IoT. Further they evaluate

the use of fog and mobile computing in collaboration

with Cloud-IoT. Fog and Cloud have been compared

using metrics like processing delay, processing cost,

processing capability and task length. An entire test

bed was used to simulate the scenarios and analyse the

results. Observations have been that Fog does reduce

processing delay to a signiﬁcant level however there

are limits to the efﬁciency. With their research they

have identiﬁed the future areas of ﬁnding the synergy

in the collaboration, scalability - horizontal and verti-

cal, and collaboration with mobile computing.

As large amounts of data is gathered even in

sub-second intervals (connected vehicle and wind

power) the existing architecture and infrastructure is

not enough to take the load of transmitting the data to

core cloud and analysing the same. One of the key re-

search for integration for all of the IoT, cloud/edge is

presented in the (Munoz and et al, 2018) which pro-

poses software deﬁned networking to arrive at the best

possible distribution of load across cloud or edge. It

builds an orchestration architecture among the com-

ponents. The authors suggest the distribution of an-

alytics between the core Cloud and the Edge of the

network to efﬁciently utilize network resources and

enable the deployment of dynamic and efﬁcient IoT

services. However, while this technique demonstrates

success, it does not leverage the potential beneﬁts of

Fog and Mobile computing architectures alongside

Cloud and IoT. A collaborative and uniﬁed architec-

ture will be able to achieve more dynamic and efﬁ-

cient analytics by distribution of analytics and efﬁ-

cient use of the network resources.

The TABLE 3 summarises the challenges/issues

in IoT, Cloud, Fog and Mobile computing integration.

Every IoT software system is characterised by

elements of compute, communication, sense/collect

data, store data and user input/output, which is de-

scribed in TABLE 4.

4 COLLABORATIVE

COMPUTING PARADIGMS -

METHOD OF SYSTEM

SOFTWARE ARCHITECTURE

To provide a method to overcome the challenges iden-

tiﬁed in Computing Paradigms and IoT, a three-prong

approach is proposed.

• First, IoT solutions need to be built on an architec-

ture of infrastructure which can enable and create

opportunities for enabling the dynamism and ex-

pansion of services

• Second, access of the components of the system

– both infrastructure and software systems – will

need to be built on well-established standards that

enable ease of operation and interoperability

• Third, on such an infrastructure software system

are to be built which can bring the functional dy-

namism

Considering that the typical system architectures

involve sensing one or more parameters and trans-

mitting to the edge or cloud processing; the lay-

ered or hierarchical approach is the ﬁrst one which

will need to change. With a notion that each of the

computing paradigms, in their heterogeneity, when

made available for a use case in a collaborative man-

ner, have the capability to enhance the system com-

pletely. Here the proposed solution is to have a com-

pletely interconnected system which has all comput-

ing paradigms accessible to the other – a total in-

terconnection which can enable a dynamic interplay.

MODELSWARD 2024 - 12th International Conference on Model-Based Software and Systems Engineering

300

Table 3: Challenges & Issues in IoT, Cloud, Fog, & Mobile Computing Integration.

No. Challenge/Issue Description

1 Data Volume, frequency and

Processing

Exponential increase in devices on the ﬁeld lead to large volume of

data collected at high frequencies. Data so collected requires adequate

processing, storage and analysis capabilities (Carla and et al., 2019).

2 Software System Architec-

ture

Developing a software system architecture that enhances the the com-

puting capabilities needed for the dynamic IoT environments (Carla and

et al., 2019).

3 Data Fusion and Dynamic

Computing

In dynamic IoT environments, heterogeneous data is collected using

numerous sensors. Fusion of such data to form relevant pieces of infor-

mation for further processing.

4 Seamless and Multi-

interface Connectivity

End-to-end connectivity with every IoT device is required for data col-

lection, device upgrades, and management. Such connectivity has to be

maintained through out the life of the system (Carla and et al., 2019).

5 Fog/Edge Computing and

Interplay with Cloud

Computing paradigms like Edge/Fog are designed to reduce the pro-

cessing delay and enable faster response times as close to the source of

data and users. Various algorithms have been explored for interplay and

federation of Fog/Edge and Cloud, however a deeper exploration of the

techniques is required (Bonomi and et al, 2012).

6 Choice of Computing

Paradigms

In practice the decision for choice of computing is largely done at de-

sign time. Orchestration policies and algorithms have been developed,

however they lack the dynamism required for IoT environments (Gio-

vanni and et al., 2019).

7 Uniﬁed Model and Resource

Allocation

Function as a service mechanism has been explored to enable a Mobile-

Edge-Cloud continuum. However, orchestration policies and strategies

for efﬁcient resource utilisation are not explored fully. (Baresi and et al.,

2019).

8 Integration and Collabora-

tion

While integration of Cloud, Fog, and IoT has been considered along

with Mobile computing, it needs further exploration, especially to bring

synergies among the participating components and scalability (Moham-

mad and et al, 2018).

Figure 1: Collaborative computing paradigms - systems architecture.

Collaborative Computing Paradigms: A Software Systems Architecture for Dynamic IoT Environments

301

Table 4: IoT Software Systems Architecture Characteristics.

IoT Architecture

Characteristics

Description

Compute Processing the data collected for variety of use cases. Processing is accom-

plished on one or more computing paradigms

Communicate Data and commands across the system over multiple connection interfaces and

protocols.

Sense/Collect Data Sensors and devices sense and collect data and use the available communica-

tion interfaces and protocols to transmit the data.

Store Data Collected data (raw), computed parameters and processed data is stored in the

system for use at various time intervals

User IO With a user being an element of the architecture, the system ultimately is re-

quired to serve the user and is required to provide a mechanism to interact.

Each of the paradigms then will be able to collabo-

rate with the other and in event of issue with non-

availability can use alternative approaches to get the

functionality completed. An MiC can collaborate

with CC without having to go through the other lay-

ers. Example: In case of an alert for a ﬁre safety sys-

tem the edge computing paradigm can itself send in

the required notiﬁcation (via email, an SMS or a push

message sent to a web portal or a mobile device).

With the interplay enabled by interconnection, the

processing of data can be dynamically distributed

across other paradigms. Such a distribution can be

done using well established mechanisms and algo-

rithms or using machine learning. This can enable

the ﬂuidity of computing where allocation of tasks is

dynamic and assignment is based on suitability and

availability of the computing paradigm in that con-

text. Let’s take a case when a model created using ma-

chine learning and large data can only be executed us-

ing cloud server thus, any model updates shall require

the cloud; however continuing functional use case can

be accomplished by other available paradigms. An

edge device can utilise the computing capability of

Edge computing as well as distribute the part to cloud

to ensure that the system is always functional. Ex-

ample: Model to control the air-conditioning may be

computed by the cloud; thus the data required for that

shall be transmitted to the cloud; while the control of

the switching of the air-conditioners can be done at

the edge using mist computing.

Data transmission among the participating

paradigms will become necessary for ensuring that

the participating paradigm has all the necessary data

and information to act on that data. Thus, input data

and processed information shall need to be stored.

Overall requirement of most systems shall be to have

a data store for maintaining the history of collected

input data and processed information which can be

achieved by using cloud storage, in the long term,

while during the functioning of the system the data

may be distributed and at frequent intervals the same

is then collected at a central location. Example:

Alerts of speeding for the driver may be generated

by using edge computing on the vehicle gateway,

and the data can be sent – opportunistically – to the

cloud later. Such data at the cloud can further be used

for processing of all alerts over a period of time to

ascertain a pattern of speeding alerts for that driver.

As and when required, each of the computing

paradigms can be augmented by addition of more in-

frastructure thus providing scalability and extendabil-

ity for each paradigm. This can be accomplished dy-

namically or as per the requirement. Example: An

increase in the amount of temperature sensors in the

data centers, the edge computing may need to be scale

the processing capability at the edge.

With the foundation of a systems architecture

of collaborative nature, the software can exploit the

available computing paradigms based on the extent of

dynamism it can support like in distributing the tasks

across paradigms.

Characteristics of the Collaborative computing

paradigms software system architecture are detailed

in TABLE 5. Each of the challenges and issues listed

in the TABLE 3 are mapped to the proposed Collabo-

rative characteristics.

5 COLLABORATIVE

COMPUTING PARADIGMS -

APPLICATIONS AND

USE-CASES

In this section two applications are discussed - Data

Center Cooling System and Vehicle and Driver Be-

haviour System. Each application presented here

is depicted using the Software Systems Architecture

without and with the proposed collaborative comput-

ing approach. A comparison, based on characteristics

MODELSWARD 2024 - 12th International Conference on Model-Based Software and Systems Engineering

302

Table 5: Collaborative Computing Paradigms- Software Systems Architecture Characteristics.

Collaborative Char-

acteristics

Description Maps

Table

Inter-connection and

interplay

Overcome the limitations imposed by layered approach by intercon-

nection and interplay between smart devices and various computing

paradigms.

5,6,7,8

Dynamic distribution

of data processing

When not bounded by strict layering and computing distribution either

at design time or by a ﬁxed algorithm, the dynamic distribution provides

increased ﬂexibility and efﬁcient distribution. With such approach the

overall available computing capacity is increased.

1, 3

Fluidity of comput-

ing across paradigms

Allocation of tasks is dynamic and assignment is based on the suitability

of a paradigm, in the context, based on the requirements and available

resources at that time.

6,7,8

Storage and data

management

across participat-

ing paradigms

All participating paradigms will require the necessary data for process-

ing and thus, the data will be distributed across paradigms. This data,

both input data and processed data, will need to be collected and trans-

mitted to a single store of data, typically on Cloud storage.

Scalability and ex-

tendability of the ar-

chitecture

The proposed architecture is designed to be scalable and extendable.

It supports the integration of newer devices seamlessly, allowing the

architecture to grow and accommodate additional components to moni-

tor subsystems, analyze data for predictive maintenance, and ensure the

scalability and reliability of the overall system.

2,4

of Collaborative computing, identiﬁed in TABLE 5,

is provided for each application.

5.1 Data Center Cooling Systems

Data centers have a requirement to maintain the tem-

perature for optimal functioning of the servers and

other network equipment. A typical installation has

servers and equipment installed in racks and data cen-

ters have multiple racks. Maintaining of the tem-

perature in the acceptable range (typically 22 to 25

deg Celsius) is accomplished using air conditioners

(ACs). At all times the goal is to provide optimal

cooling with lower consumption of electricity. With

this goal in mind, temperature and electrical con-

sumption are monitored on regular basis.

5.1.1 Case (1) AC with IoT (Classical Layered

Approach) (Kaushik and Naik, 2023)

IoT devices for sensing temperature and humidity are

installed on server racks and in the data center at the

desired locations. The energy consumption for each

of the ACs is monitored using smart energy meters

and the remote control for air conditioners is made

possible by smart devices. In this case we consider the

ductless ACs. Temperature, humidity and energy con-

sumption is collected at regular intervals (typically 10

to 30 min) along with time stamps for each of the

readings. This data is then transmitted over a net-

work interface, wired and/or wireless, to the Cloud

server which pre-processes the data and stores in a

database. Typically same or another server performs

the computation, by accessing the data over the net-

work from the cloud data store. This computation and

processing is performed to build a cooling model (a

Machine Learning (ML) algorithm). Based on such

a cooling model, commands for switching the ACs

are transmitted over the network to the smart device.

Outcome of such a system is results in the switching

of ACs, the operating hours life of the system is ex-

tended and electricity consumption is reduced. Num-

ber of sensors may be added to improve the cooling

model which will increase the Cloud server load with

additional data collection, storage and processing.

In regular operations the overall ACs can be mon-

itored for healthy operations and any faults may be

communicated to the cloud server and processed to

create a series of alerts and notiﬁcations which can be

sent over channels like email, SMS or notiﬁcations,

using web-sockets, on the portal.

With the switching of ACs controlled by the cloud

servers, in case of network connectivity issues, the

commands for switching the ACs will not reach the

smart device and the operations are at risk. While fail

safe may be built for ACs to continue operation in

such a scenario; overall operations will lead to inef-

ﬁcient consumption of electricity by the ACs as the

switching will be affected. In such a scenario alerts or

notiﬁcations also will not be generated and thus, the

Collaborative Computing Paradigms: A Software Systems Architecture for Dynamic IoT Environments

303

only way to monitor the system shall be manual by

physical presence at the the data center.

5.1.2 Case (2) AC with IoT (Collaborative

Computing)

With the IoT devices for sensing and control in place,

same as in Case (1), the smart device, used for switch-

ing is made capable of storing the schedule of switch-

ing of the ACs in an autonomous manner. In this case

the computation of the cooling model can be done

on the cloud server based on the data collected and

the switching duty cycle can be communicated to the

smart device which shall operate the ACs in an au-

tonomous manner. Thus, reducing the need for the

cloud server to be actually responsible for the control

of switching of the ACs. Further, an edge gateway,

which is capable of computation and control for the

entire data center, can be installed, which can reduce

the computation and control burden from the cloud

servers. This will lead to a system where the edge de-

vices take over the prime responsibility, Cloud server

is freed up to compute the model. Cloud server’s

resources, which otherwise were responsible for the

switching, can now be used to build a better model

and even compare with the data from other data cen-

ters. Edge device and Cloud server can establish an

interplay and collaboration per the increase in the data

center load or amount of data collected as more sen-

sors get added for betterment of the cooling system.

A mobile phone, with an app, can be used to remotely

monitor the system and receive alerts. Figure 2 de-

picts the proposed Collaborative software system ar-

chitecture for Data Center Cooling System.

In the event of network connectivity issues with

the cloud server, the entire operation can still continu-

ally smoothly with the autonomy of the smart devices

and the edge gateway. In event of fault, the alerts and

notiﬁcation can be generated and transmitted by the

edge gateway based on the interfaces available on the

edge gateway.

Case 1 and 2 of the Data Center Cooling System,

are evaluated based on the proposed characteristics of

Collaborative computing paradigms and depicted in

TABLES 6 and 7. It can be observed from TABLES

6 and 7 that Case 2 leverages the Collaborative Com-

puting for achieving better functionality.

5.2 Vehicle and Driver Behaviour

In order to improve safety of the vehicles and drivers

on the road, numerous solutions based on IoT have

been deployed, which are based on the data collected

from the On Board Diagnostics (OBD) port. Vehicle

parameters like speed, engine rpm, odometer, coolant

Figure 2: IoTD, CC, EC, MC: An architecture for Data Cen-

ter Cooling System.

temperature, fuel level, fuel consumption, gear posi-

tion are collected at frequent internals, from the Elec-

tronic Control Units (ECU) and are used for driver

and vehicle behaviour analysis. Fault codes detected

on the OBD can provide further information on need

for service for the vehicle. Data is also collected from

other sensors and smart devices like GPS, video from

dashcam that provide vital information on location

and environment. Based on these parameters further

attributes can be computed like hard braking, accel-

eration, deceleration, adherence to speed limits which

are useful for providing feedback to the drivers. Fault

codes and other engine details can provide informa-

tion on the running and ﬁtness of the vehicle. Data

collected from the vehicle can also be processed for

predictive analytics for preventive maintenance.

5.2.1 Case (1) Vehicle OBD with IoT (Classical

IoT and a Layered Approach)

An OBD device and a dashcam, installed on the vehi-

cle, collects the data and transmits it to the Cloud for

further processing. In some cases there may be a ve-

hicle gateway that may collate the data, preprocess it

- typically add time stamps, create packets of relevant

corelatable data, example add location information,

and organise it in ﬁles - before the data is sent to the

cloud. Cloud server processes the data and provides

the feedback in terms of alerts like excess speed, hard

braking and acceleration to the driver and those are

shown on the vehicle gateway or on the mobile phone

of the driver. A mobile app, for the driver, provides

the necessary processed information (received from

the cloud server) in the form of information like driver

score or alerts like hard-braking or acceleration.

In this case the complete processing done by the

cloud servers, in event of network connectivity issues,

the processed information and alerts will not be avail-

able for the driver. Additional features may be built

on the mobile phone app which can perform minimal

MODELSWARD 2024 - 12th International Conference on Model-Based Software and Systems Engineering

304

Table 6: Collaborative computing for Data Center Cooling System.

Collaborative Computing Characteristics Case 1 Case 2

Interconnection and interplay NO YES

Dynamic distribution of data processing NO YES

Fluidity of computing across paradigms NO YES

Storage and data management across participating paradigms NO YES

Scalability and extendability of the architecture NO YES

Table 7: Application: Data Center Cooling System.

Design of cooling

system

Temperature Control Electricity Optimisa-

tion

Scalability

Without IoT YES. Localised None None

With IoT YES. Cloud Con-

trolled

YES. Cloud Con-

trolled

YES. Need more

computing on Cloud

IoT with Collabora-

tive

YES. Edge Con-

troller/Cloud guided

YES. Edge Con-

troller/Cloud guided

YES. Cloud re-

sources optimised

processing to alert the driver. Example: App on mo-

bile phone may be able to generate hard braking or

acceleration events to alert the driver.

5.2.2 Case (2) Vehicle OBD with IoT

(Collaborative Computing)

In addition to the set-up as mentioned in Case (1),

a vehicle gateway with better computing capability

(Edge computing) is installed. The vehicle gateway

typically will be equipped with multiple interfaces to

connect with devices on board like the OBD, dash-

cam and interface like Bluetooth or USB to connect to

other devices like Mobile Phones or Laptop. The ve-

hicle gateway can communicate with the Cloud server

over a wireless connection. As in Case (1), the mobile

phone of the driver can also participate in the overall

solution.

With such an architecture, the vehicle gateway, the

mobile phone and Cloud server can collaborate for

preprocessing and processing the data. Example the

hard braking, acceleration alerts can be generated on

the mobile phone or the vehicle gateway, for speeding

alerts the vehicle gateway can connect with the Cloud

server to ascertain the speed for the particular road

segment and compare with the actual speed of the ve-

hicle, a collaborative method. Since the connectivity

exists among all participating computing paradigms,

depending on the load they can dynamically share the

computing resources.

With all the computing paradigms (CC, EC and

MC) working in a collaborative way, speciﬁc com-

puting load can be shared with the EC or MC, thus

increasing the overall capacity of the computing in-

frastructure. Figure 3 depicts the proposed Collabo-

rative software system architecture.

Case 1 and 2 of Vehicle and Driver Behaviour,

Table 8: Collaborative computing for Vehicle and Driver

Behaviour.

Collaborative Characteris-

tics

Case 1 Case 2

Interconnection and inter-

play

PARTIAL YES

Dynamic distribution of

data processing

NO YES

Fluidity of computing

across paradigms

NO YES

Storage and data manage-

ment across participating

paradigms

PARTIAL YES

Scalability and extendabil-

ity of the architecture

PARTIAL YES

are evaluated based on the proposed characteristics

of Collaborative computing paradigm and depicted in

TABLE 8. As noted in description of earlier applica-

tion, Collaborative Computing provides better func-

tionality for identiﬁed characteristics.

6 CONCLUSION

Based on the challenges and issues identiﬁed for com-

puting paradigms and IoT, a collaborative computing

system architecture is developed. The characteristics

for the developed architecture have been identiﬁed

and are mapped to the challenges and issues identi-

ﬁed. It is clear from the mapping that the architecture

developed is capable of addressing the challenges and

issues. The proposed architecture is validated through

it’s application for vehicle and driver behaviour, and

data center cooling systems. A comparison has been

done for with and without the collaborative comput-

Collaborative Computing Paradigms: A Software Systems Architecture for Dynamic IoT Environments

305

Figure 3: IoTD, FC/EC, MC, CC: An architecture for Vehi-

cle and Driver behaviour.

ing architecture, For both the applications it is evi-

dent that this method adds dynamism, system reliabil-

ity and higher utilisation of resources. It has demon-

strated better capability to handle dynamic environ-

ment like IoT and better capability to handle uncer-

tainty of connectivity issues. This architecture en-

hances the capability of computing paradigms char-

acteristics when multiple paradigms are employed for

a real-world system. With the capability to execute

the tasks at any of the available computing paradigms

or across multiple paradigms in a collaborative way,

the overall higher system efﬁciency and utilisation of

resources can be achieved.

7 FUTURE WORK

Further to this, the need is to experiment and apply for

a variety of applications by focusing on the charac-

teristics of interplay, dynamism, and ﬂuidity of com-

puting. With the collaborative method, the data stor-

age across paradigms of preprocessed and processed

data will need speciﬁc algorithms to collect the data

at one location say Cloud. With the collaborative

method, the distribution mechanisms of tasks across

paradigms will need to be further studied along with

effectiveness of AI and ML on dynamic environment

like IoT.

REFERENCES

Baresi, L. and et al. (2019). A uniﬁed model for the mobile-

edge-cloud continuum. In ACM Transaction on Inter-

net Technology. ACM.

Bonomi, F. and et al (2012). Fog computing and its role in

the internet of things. In IEEE. IEEE.

Cai, Q. and et al (2023). Collaboration of heterogeneous

edge computing paradigms: How to ﬁll the gap be-

tween theory and practice. In IEEE Wireless Commu-

nications. IEEE.

Carla, M. and et al. (2019). A comprehensive survey on fog

computing: State-of-the-art and research challenges.

In IEEE Communications Surveys & Tutorials. IEEE.

Donno, M. D. and et al. (2019). Foundations and evolution

of modern computing paradigms: Cloud, iot, edge,

and fog. In IEEE Access, vol. 7, pp. 150936-150948.

IEEE.

Escobar, L. and et al (2022). In-depth analysis and open

challenges of mist computing. In Journal of Cloud

Computing. Springeropen.

Fern

andez, C. M. and et al (2018). An edge computing ar-

chitecture in the internet of things. In IEEE 21st Inter-

national Symposium on Real-Time Distributed Com-

puting (ISORC). IEEE.

Garc

es, L. and et al (2021). Three decades of software ref-

erence architectures: A systematic mapping study. In

Journal of Systems and Software. Elsevier.

Giovanni, M. and et al. (2019). Enabling workload en-

gineering in edge, fog and cloud computing through

openstack-based middleware. In ACM Transaction on

Internet Technology. ACM.

Iorga, M. and et al. (2018). Fog computing conceptual

model. In NIST Standards and Technology. NIST.

Kaushik, K. and Naik, V. (2023). Making ductless-split

cooling systems energy efﬁcient using iot. In Interna-

tional Conference on Communication Systems. COM-

SNETS.

Lewandowski, T. and et al (2020). A software architecture

to enable self-organizing, collaborative iot ressource

networks. In Fifth International Conference on Fog

and Mobile Edge Computing (FMEC). IEEE.

Mahmoudi, C. and et al. (2018). Formal deﬁnition of edge

computing: An emphasis on mobile cloud and iot

composition. In Third International Conference on

Fog and Mobile Edge Computing (FMEC). IEEE.

Malekian, R. and et al. (2015). Design and implementation

of a wireless obd ii ﬂeet management system. In IEEE

Sensors Journal, vol. 17, no. 4, pp. 1154-1164. IEEE.

Mohammad, A. and et al (2018). Fog computing archi-

tecture, evaluation, and future research directions. In

IEEE Communications Magazine. IEEE.

Munoz, R. and et al (2018). Integration of iot, transport sdn,

and edge/cloud computing for dynamic distribution of

iot analytics and efﬁcient use of network resources. In

IEEE Journal of Lightwave Technology. IEEE.

Nascimento, M. G. D. and et al (2023). An architec-

ture to support the development of collaborative sys-

tems in iot context. In IEEE International Conference

on Computer Supported Cooperative Work in Design

(CSCWD). IEEE.

Peter, M. and Timothy, G. (2011). The nist deﬁnition of

cloud computing. In NIST Standards and Technology.

NIST.

urker, G. F. and et al (2016). Survey of smartphone appli-

cations based on obd-ii for intelligent transportation

systems. In International Journal of Engg Research

and Applications. IJERA.

Vasconcelos., D. R. and et al. (2019). Cloud, fog, or mist

in iot? that is the question. In ACM Transaction on

Internet Technology. ACM.

MODELSWARD 2024 - 12th International Conference on Model-Based Software and Systems Engineering

306