ArticlePDF Available

Coordination Protocol and Admission Control for Distributed Services in System-of-Systems With Real-Time Requirements

Authors:

Abstract and Figures

System-of-Systems (SoS) offer unprecedented potential for new types of emerging services, which significantly exceed the capabilities of the constituting systems. SoS in safety-critical domains (e.g., medical applications, smart grid, disaster recovery, defense) are prominent examples, but they have stringent real-time and reliability requirements. Therefore, a suitable temporal and spatial allocation of resources is required both within each constituent system and in the wide area networks between them. This paper introduces an algorithm for admission control and resources’ allocation, which considers these requirements and the autonomy of the constituent systems. To simulate a realistic admission control and resources’ allocation process of a typical SoS network, a simulated case study with eight constituent systems, six services, and twenty-five processes/requests is developed. The suggested admission control and resources’ allocation process’s performance is measured in terms of gain in the execution time and blockage probability. A sensitivity analysis is carried out to evaluate the influence of the number of constituent systems and the number of services sought by the received processes/requests on the efficacy of the proposed process. The results show that the proposed admission control and resources’ allocation process have very low blockage probability, high gain in the execution time, and high resources’ utilization.
Content may be subject to copyright.
Date of publication xxxx 00, 0000, date of current version xxxx 00, 0000.
Digital Object Identifier 10.1109/ACCESS.2017.DOI
Coordination Protocol and Admission
Control for Distributed Services in
System-of-Systems with Real-time
Requirements
DHIAH EL DIEHN I. ABOU-TAIR1, ALA’ KHALIFEH1, (Member, IEEE), Sameer
Al-Dahidi2,SAHEL ALOUNEH1,3, and ROMAN OBERMAISSER4
1School of Electrical Engineering and Information Technology, German Jordanian University, Amman, Jordan
2School of Applied Technical Sciences, German Jordanian University, Amman, Jordan
3College of Engineering, Al Ain University, Abu Dhabi, U.A.E.
4Chair for Embedded Systems, University of Siegen, Siegen, Germany
Corresponding author: Dhiah el Diehn I. Abou-Tair (e-mail: dhiah.aboutair@gju.edu.jo).
This work has been supported in part by the European research project FRACTAL under the Grant Agreement ID 877056 and the DFG
project with the grant number OB 384-11-1.
ABSTRACT System-of-Systems (SoS) offer unprecedented potential for new types of emerging services,
which significantly exceed the capabilities of the constituting systems. SoS in safety-critical domains
(e.g., medical applications, smart grid, disaster recovery, defense) are prominent examples, but they have
stringent real-time and reliability requirements. Therefore, a suitable temporal and spatial allocation of
resources is required both within each constituent system and in the wide area networks between them.
This paper introduces an algorithm for admission control and resources’ allocation, which considers these
requirements and the autonomy of the constituent systems. To simulate a realistic admission control and
resources’ allocation process of a typical SoS network, a simulated case study with eight constituent
systems, six services, and twenty-five processes/requests is developed. The suggested admission control and
resources’ allocation process’s performance is measured in terms of gain in the execution time and blockage
probability. A sensitivity analysis is carried out to evaluate the influence of the number of constituent
systems and the number of services sought by the received processes/requests on the efficacy of the proposed
process. The results show that the proposed admission control and resources’ allocation process have very
low blockage probability, high gain in the execution time, and high resources’ utilization.
INDEX TERMS System-of-Systems, admission control, resources’ allocation, distributed systems, real-
time requirements
I. INTRODUCTION
System-of-Systems (SoS) have recently gained popularity
among the research community, especially with the advent of
several emerging and disruptive technologies like the Internet
of Things (IoT) [1], cloud computing [2], [3], big data [4],
Artificial Intelligence (AI) [5], etc. Therefore, a new era
of applications and services driven by a mixture of these
technologies, has been introduced to cope up with our infor-
mation age, thus research around SoS network architecture
and paradigm becomes more in focus. Electronic health care
services [6], remote security and monitoring [7], precision
agriculture [8], aviation [9] and industrial automation [10]
are some examples of these technology driven applications
and services which can be effectively integrated and im-
plemented in the context of SoS. However, to establish a
reliable and effective SoS infrastructure for these services,
the conventional Information Technology (IT) infrastructure
that solely depends on the resources available at one single
network, should be revamped in a way to support the diverse
resources needed by these services. Thus, the SoS infras-
tructure requires the integration and collaboration of sev-
eral network service providers. For instance, several safety-
relevant applications and scenarios often have stringent re-
quirements such as reliability, security, high availability, and
VOLUME 4, 2016 1
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
real-time constraints. Capitalizing the resources needed to
meet the requirements of these applications and services in
one network is very challenging and, in some case, impos-
sible, especially when considering the cost, effectiveness,
scalability, reliability, and efficiency aspects. Motivated by
the above discussion, SoS architecture was proposed as an
efficient and cost effective solution to these challenges, which
in turn will pave the way toward actualizing these services
in an affordable manner, taking into account the application
requirements and constraints, thus ensuring high quality SoS
services. The SoS initial system architecture was coined in
Maier [11] sentimental paper, where the author defined the
main taxonomies associated with SoS, along with the princi-
ple architecture. Maier emphasized that the SoS architecture
is not physical but rather logical, where a set of standard
components communicate with each other, to accomplish a
certain functionality in a collaborative manner. The funda-
mental benefit of the SoS architecture lies in the fact that
it utilizes the available resources for different systems, that
work together as one complete system thus being capable of
providing complex functions, which cannot be achieved by
the independent systems.
This paper uses the definition of SoS as introduced in [12],
[13], dependent and which are networked together for a
period of time to establish emerging services and satisfy
requirements that cannot be met by any CS in isolation.
Cyberphysical SoS are time-sensitive and include not only
the computer system, buty also a controlled object and hu-
mans [12]. An example of such a system is an intelligent
transportation system with vehicle-to-vehicle communica-
tion and vehicle-to-infrastructure communication. Another
example is a healthcare SoS where patient monitoring and
treatment services are dynamically established with patients,
medical doctors, hospitals, data centers and care takers.
For a given application to be deployed in an SoS network
architecture, the application is initiated at a CS, which is
often named as the initiator CS as will be explained in the
SoS section of this paper (section III). The initiator-CS is
responsible for resources provision and coordination between
the SoS CSs, which involves (1) identifying CSs that can
provide the services, (2) optimizing the use of the services
driven by extra-functional properties, (3) performing admis-
sion control for service-provision, (4) reserving resources for
provided services and (5) recursively using services from
subcontracted CSs to realize service-provisions [14].
This paper focuses on the admission control challenge in
SoS network architecture, which is very crucial for effective
and efficient service provisioning. Admission control in SoS
ensures that the required resources demanded by the user
request are fulfilled taking into consideration many factors
such as: the service real-time provisioning, priority, relia-
bility, cost, scalability, security, etc. In our previous works,
[15], [16], an SoS model and architecture were introduced. In
[17], [18], we focused on the security and network resource
reservation between different CSs within the SoS utilizing
the Multiprotocol Label Switching (MPLS) paradigm. In
[19], the SoS network architecture taking into account the
services’ provisioning concept for SoS have been described.
Furthermore, a preliminary description of a proposed dis-
tributed admission control algorithm has been described in
[20]. This paper expands our previous work in the field of
SoS, mainly the one presented in [20] by expanding the
proposed admission control algorithm, where a more detailed
description is provided, in addition to presenting a detailed
simulation and performance evaluation of the proposed al-
gorithm for SoS network architecture, which addresses the
problem of performing admission control for service provi-
sioning utilizing the distributed SoS resources. Further, the
paper provides a more holistic view on the network archi-
tecture, considering an SoS paradigm, where the resources
must be first checked from availability point of view, then
reserved and utilized by the requesting service. The proposed
algorithm has been evaluated by simulation, which showed
the effectiveness of the proposed algorithm. In summary, the
original contributions of this work can be summarized as
follows:
The development of an admission control and resources’
allocation algorithm that considers the application under
consideration requirements and the autonomy of the
CSs.
The verification of the proposed algorithm with respect
to a simulated case study, properly designed to mimic
a realistic admission control and resources’ allocation
process of a typical SoS network.
The evaluation of the proposed algorithm’s effectiveness
by computing various performance metrics from the
literature, such as the gain in the execution time and the
blockage probability.
The investigation of the influence of the number of
CSs and the number of services sought by the received
processes/requests on the efficacy of the proposed algo-
rithm by a proper sensitivity analysis.
Furthermore, most of the published works in the literature
propose admission control algorithms for one system, while
this work focuses on proposing an admission control algo-
rithm for distributed services in SoS with real-time require-
ments.
The rest of the paper is organized as follows: Section II
summarizes the state of the art papers. Section III describes
the SoS network architecture. Section IV discusses the pro-
posed admission control and resource allocation algorithm.
The performance evaluation of the proposed algorithm is
presented in Section V. Finally, the paper is concluded in
Section VI and future recommendations are stated.
II. LITERATURE REVIEW
The SoS literature covers a wide range of applications, types,
and research issues. We limit ourselves to the admission
control and scheduling part. There has been numerous works
in the field SoS, however, there is a little work that focuses
on and solve issues related to the coordinated and distributed
admission control and scheduling for SoS networks.
2VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
For example, the authors in [21] tackled the issue of
scheduling large size of dependent and independent task
requests that reach cloud systems. The work proposed a
heuristic scheduling mechanism to reduce the time of large
requests scheduling. The results obtained presented an im-
proved performance outcomes when compared to other cur-
rent solutions in terms of speed-up, make-span, and length ra-
tio scheduling. The dynamic request scheduling and resource
allocation within consentient systems are considered in [22],
[23].
The work in [24] presented a data processing platform
named Nephele that dynamically assigns resources in cloud
systems for executions and scheduling tasks. The proposed
platform does not consider Quality of Service (QoS) require-
ments and end-to-end deadlines in the scheduling process.
Furthermore, reliability and the network inconsistency be-
tween the constituent systems are not considered.
A literature survey of resources’ scheduling algorithms
for cloud computing systems was discussed and analyzed
in [25]. The survey and analysis considered the grouping
of resources, evolution of resources scheduling, and differ-
ent available scheduling algorithms and the associated QoS
constraints. The authors indicated that it is still not an easy
task to discover the best planning and mapping of workloads
capabilities and resources without having effective mecha-
nisms for resource provisioning. Also, it can be observed
from the conducted literature review and surveys that most
of the existing works consider only scheduling requests and
resource allocations on one and only CS, while the require-
ments for multiple CSs of diverse requests are not taken into
consideration. Additionally, the vibrant nature of resources
availability and requests are not addressed carefully.
The researchers in [26] provided new awareness into the
energy sharing and low-carbon cost-effective scheduling be-
tween multiple energy systems from the systems of systems
perspective. The relation between admission control and data
analytic is also of main concern.
An aircraft fleet planning and architecture framework as-
sessment for air mobility and distribution using a system-
of- systems approach is discussed in reference [27]. The
framework provides understanding of the SoS design space
and successful deployment or optimization of UAM fleet.
The concept of emergence and its relation to the engi-
neering of SoS and Smart energy grids environmental impact
were studied in [28].
A recent work that proposes an algorithm for incremental
scheduling with real time requirements for the SoS is dis-
cussed in [29]. The authors proposed a two-level cooperating
heuristic method using a Genetic Algorithm (GA) to schedule
real-time requests incrementally by adding requests to the
time triggered SoS systems. The algorithm deploys and com-
putes the schedule for every new SoS request after its arrival
at run-time. Consequently, the computational resources and
limited communication must be shared and communicated
between different SoS requests and applications released and
communicated over time. The blocking time thus is reduced
for the shared resources; and as a result, the possible shortage
of resources for future applications can be handled. Hence,
the authors developed a new resource allocation algorithm
for more balanced blocking times of shared resources and
supports future resource requests.
The work in Roy et al. [30], proposed two Integer Lin-
ear Programming (ILP)-based schemes, namely ILP with
Explicit Time Reduced (ILP-ETR) and ILP with Non-
overlapping Constraints (ILP-NC), for optimally schedul-
ing real-time Precedence-constrained Task Graphs (PTGs)
on platforms composed of heterogeneous processing ele-
ments interconnected through a set of heterogeneous shared
buses in contrast to conventional schemes that deal with
homogeneous elements and communication channels. The
suggested schemes were shown to be realistically efficient
when tested on an automotive cruise controller case study.
In the same context of a heterogeneous distributed platform,
the work of Roy et al. [31], proposed two low-overhead
heuristic algorithms, namely Global Slack Aware Quality-
level Allocator (G-SLAQA) and Total Slack Aware Quality-
level Allocator (T-SLAQA) for optimally scheduling real-
time Directed-acyclic Task Graph (DTG) combined with
multiple quality-level tasks, with convenient computational
efforts. The proposed schemes were shown to be more ef-
fective than the traditional ILP scheme when tested in an
automotive traction controller case study. Similarly, in the
work of Roy et al. [32], the authors proposed a low-overhead
heuristic algorithm, namely the contention cognizant task
and message scheduler (CC-TMS) for optimally scheduling
real-time DTG with convenient computational efforts. The
proposed scheme was shown to be more effective than the
traditional ILP scheme when tested in an automotive traction
controller case study.
The authors in [33] illustrated the use of AI to sort space
habitation sub-systems for NASA technological groups and
to classify applicable sources of data for these sub-systems.
The authors demonstrated how AI agents can support the re-
covery and retrieval of composite information needed to feed
existing SoS analytic tools and discussed possible challenges
and future steps.
The authors in [34] discussed the problem of managing
volatile and unpredictable variations of SoS networks. The
authors proposed a dynamic reconfiguration scheme that
aims at enhancing the SoS agility to quickly respond from
failures. The proposed scheme employs estimated dynamic
programming technique to calculate the dynamic reconfig-
uration choices and decisions that can allow failed or de-
graded sub-systems to be detached and the allocation of new
resources to be changed rapidly.
III. SYSTEM-OF-SYSTEMS ARCHITECTURE
As shown in Fig. 1, a typical SoS consist of several CSs
where each CS can deliver certain services provided by its
End Systems (ESs), which can be for example a computing
device that runs a certain software application to provide
services. The ESs are connected with each other via routers.
VOLUME 4, 2016 3
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
Further, each CS contains a Constituent System Manager
(CSM) that manages the CS resources in terms of admission
control, resource allocation and scheduling.
Within this SoS architecture, the envisioned establishment
process for dependable SoS-applications as depicted in Fig. 3
is as follows:
a: SoS-Application Request and Target-CS Discovery.
The starting point is a request for an SoS-application at a CS
(henceforth called initiator-CS). As part of a graph of services
depicted in Fig. 2, the initiator-CS will require services
from other CS (henceforth called target-CS). Therefore, the
initiator-CS needs to find CS that provides these services by
using a broker (e.g., [35]).
b: Service-Offer Request.
The initiator-CS will send requests for service-offers to the
target-CS, which were named as potential providers of the
services by the broker.
c: Short-term service admission by target-CS.
Each target-CS will execute a schedulability test and in case
of success it will reply with a service-offer and a short-term
admission to the initiator-CS. The short-term admission will
be associated with an expiration time. Each target-CS can
recursively contact other CS if it needs to use other CS-
services to provide its own services.
d: Optimized Selection of Target-CS.
The initiator-CS checks the service-offers from the target-
CS by considering their available resources and matches it
with the requests’ constraints (e.g. end-to-end delay). The
initiator-CS selects for each service the winner CS(s) in an
optimized manner (e.g., cost, reliability, timing, network de-
lays, network reliability). Confirmation messages will trans-
form the short-term admission into a long-term admission.
ES
Router
CSM
Router
ES
ES ES
Constituent
System: CS
ES
Router
CSM
Router
ES
ES ES
Constituent
System: CS
ES
Router
CSM
Router
ES
ES
ES
Constituent
System: CS
ES
ES
ES
Wide Area Network
ES
Router
CSM
Router
ES
ES ES
Constituent
System: CS
ES
Router
CSM
Router
ES
ES ES
Constituent
System: CS
CSM ConstituentSystemManager
NMS NetworkManagementSystem
ES EndSystem
Router
Router
Router
FIGURE 1: System architecture with Constituent Systems
(CS), End Systems (ES) and Constituent System Manager
(CSM) [19].
Service
Service Service
Service Service
Service
Service
ServiceService
Service Service
CSMCS1
CSMCS1
CSMCS2
CSMCS6CSMCS3
CSMCS4
CSMCS5
CSMCS2
FIGURE 2: Service graph in an application of the SoS (left)
and Interaction graph between CSMs (right) [19].
ES
Router
CSM
Router
ES
ES ES
Initiator
Constituent
System
CSNS
Router
CSM
Router
ES ES ES
Target Constituent
System
ES
ES
ES
ES
Router
CSM
Router
ES
ES ES
Target Constituent
System
ES
Router
CSM
Router
ES
ES
ES
Target Constituent
System
Incremental
Resource
Reservation
Broker
FIGURE 3: Envisioned establishment process of a depend-
able SoS-application [19].
Short-term admissions without a confirmation from the ini-
tiator will expire at the target-CS.
e: Long Term Admission and Execution of SoS Application.
The selected target-CS will execute a distributed algorithm
for the incremental resource reservation of these CSs. There-
after, the SoS-application is executed based on the allocated
resources and the service contracts. Service revocations can
occur in case of resource conflicts with SoS-applications of
higher criticality.
IV. ADMISSION CONTROL AND RESOURCES
ALLOCATION
To establish the process of dependable SoS applications, a
coordination protocol needs to be defined that is responsible
for resources’ scheduling among CSs in a distributed manner.
In the proposed SoS network architecture, several requests
submitted from different ESs asking for services with diverse
4VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
requirements such as E2E delay, reliability, security, costs,
etc. need to be processed by different CSs. These requests
are generated from different CSs that may be issued for the
same services at the same time. The coordination protocol
must handle these requests dynamically and in a distributed
manner such that each CS will handle the received/generated
requests and allocate the needed resources in coordination
with the other related CSs. This allocation process optimizes
the available resources in all CSs to assure that all requests
meet their requirements. The challenge in designing the
coordination protocol results from the stringent requirements
of the architecture and the applications of the SoS. A concur-
rent, incremental, and distributed scheduling and admission
control process is required as described below.
1) Admission control: In SoS, different CSs may gen-
erate different requests with specific constraints. Ad-
mission control determines whether the requests can
be admitted and scheduled according to the available
resources, such that the requests’ requirements can be
met. To achieve that, CSs must exchange their resource
allocation and scheduling policies regularly in order
to get a global picture of the available resources and
determine if the new request can be admitted taking
into account its constraints without jeopardizing the
admitted requests.
2) Requests concurrency in mixed-critical systems: CSs
can receive multiple requests with different priorities
and criticalities at the same time competing for the
current available resources. To handle these requests,
a prioritization policy has to be implemented in all CSs
such that it handles the requests in an optimal manner
taking into account their criticalities and the resource
availability. For example, some requests should be han-
dled with stringent deadline requirements, since they
are related to safety-critical services. These requests
are considered more critical than other less stringent
requests. This process may cause different challenges
that need to be addressed such as deadlocks, request
starvation and aging.
3) Revocation of previous admission based on criticality:
High priority requests may arrive to CS without being
admitted due to the lack of the available resources.
To avoid such a case, resources revocation for lower
priority admitted requests should take place, which will
allow the admission of higher priority requests. This
resource allocation and revocation process based on
the priority should be done in coordination with all
CSs to ensure optimal resource allocation with minimal
resources wastage.
4) Handling requests with multiple constraints: Several
requests with different constraints (e.g., real-time as-
surance, reliability, security, cost, etc.) should be han-
dled properly in the resources’ optimization alloca-
tion process of different CSs. However, dealing with
different constraints increases the admission process
complexity. As such, heuristic resource allocation al-
gorithms must be proposed in all CSs to deal with the
problem efficiently.
To have an optimal distributed resource allocation between
different CSs, a distributed resource allocation protocol is
proposed. The protocol runs on all CSs of the SoS and
implements the following tasks:
1) Resource Discovery. Each CS must have up-to-date
knowledge about the available resources at its ESs
with their current status. To achieve that, a Resources
Allocation Manager (RAM) is proposed, which runs
a periodic resource discovery process to explore any
new resources that were added and also to exclude
vanished resources. While running the discovery pro-
cess, a Resources Allocation Table (RAT) should be
established which has the following main entries: CS
ID, Resources ID (RID), Resource Status (RS), Priority
Level (PL), and Process Constraints (PCons). In what
follows, a brief description of the RAT fields is pro-
vided.
Resources ID: each resource should have a unique
ID within the SoS. This ID should indicate its main
functionality and the ES that hosts the resource.
Moreover, depending on the CS that is hosting
the resource, each resource will have different
processing time. This can be represented using the
Resource Processing Time (RPT) entry.
Resource Status: the RAM should be aware of the
status of all resources, which is represented by a
tuple of three fields; the Resource Status which
can be either Busy (B) or Free (F), the Process
ID that is occupying the resource, and the time
slot on which the resource will be busy (Busy Slot
(BS)). For example, if the resource R1in CS0is
occupied by process ID (P0), during the period
from [0-75] ms, then the RS of the resource R1
inside CS0RAT will be as follows: B:P0:BS[0-
75]. It is important to mention that to maximize
the resources utilization, the entries of the RAT of
each CS should be regularly updated whenever a
new process is admitted or served.
Priority Level: once a process utilizes resources,
its priority level should be saved. A mechanism
for determining the process priorities should be
implemented.
Process Constraints: all the process constraints
such as E2E delay, fault tolerance, security, reli-
ability, etc. are stored in this field. In addition, this
field contains the set of the needed resources by
the process at each CS.
CSs Route: this entry shows the sequence of re-
maining CSs (including the one that is admitted
to), that the process will follow to get all the
needed resources and services.
Finally, it is important to emphasize that this table
VOLUME 4, 2016 5
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
is frequently exchanged with the neighbouring CSs
within the SoS, so all the CSs will have a global
vision for all the available resources in different CSs.
Once the resource discovery process is performed, the
path of the CSs that have the needed resources can be
identified, which will be used in the path reservation
process utilizing the Resource Reservation Protocol
(RSVP).
2) Resource Reservation. In order to serve multiple pro-
cesses with different priorities and requirements, a
resource reservation process will take place. To achieve
that, a Resource Reservation Protocol (RRP) is pro-
posed. The RRP has the following main functionalities:
Process Pre-admission: When a process arrives to
the SoS, the RRP will admit it temporarily till it
checks whether it can fulfil its constraints. The
RRP will save the received requests in a temporary
queue, extract its requirements and check with
other SoS for resource availability.
Path Determination: the RRP will consult the CSM
of the CS in order to determine all possible paths
toward the next CS. Different paths can be utilized
to choose the one that can fulfil the process con-
straints.
Resources Reservation: Once the path of the next
CS is determined, the RRP will send a Resource
Reservation Request (RRR) with the following
fields: The Process ID, which defines the ID of
the process that is asking for a resource. The
Source CS ID is the ID of the CS that the process
belongs to. The Resource ID is the ID of the
resource required by the process and the Priority
Level defines the process priority level to assess
its criticality. The constraints define the process
constraints in terms of E2E delay, fault tolerance,
security and reliability.
Process Admission. The RRR will be sent to all
the required CSs needed by the process. If all
the needed CSs can fulfill the process require-
ments, then the needed resources in all CSs will
be reserved, and the process is moved from the
pre-admission queue in the receiving CS to the
admitted queue. All the CSs will update their RAT
to reflect the new admitted processes. However,
if the reservation process was unable to fulfill the
process requirements and constraints, then another
admission process called Priority-Based Admis-
sion (PBA) will take place, which is used as a
mitigation procedure for the potential failure of the
normal admission process. This may happen if the
admission process was unable to admit a process
since other processes are occupying the available
resources. In this case, the process priority should
be considered. The flow chart and the unified
modeling language (UML) sequence diagram of
the admission control algorithm are depicted in
Fig. IV, and Fig. IV, respectively. Furthermore,
the admission control computational complexity
is O(RsN), where Rsis the maximum resources
requested by the processes, and Nis the num-
ber of CSs. This complexity corresponds to the
worst case scenario where the initiator CSM has to
check with all CSs for all requested resources by
the process. Finally, the spatial complexity which
corresponds to the storage needed to store the
CSs RAT is O(TsN), where Tsis the maximum
possible RAT table size (which corresponds to the
CS that has the most resources. It is apparent that
both complexities are polynomial which makes the
algorithm feasible from implementation point of
view.
A. DEMONSTRATION EXAMPLE
The following example aims at elaborating the proposed
process admission and resources reservation protocol, which
also highlights the various challenges and requirements that
need to be taken into consideration while developing the
admission control and resource allocation algorithms.
Fig. 6 shows SoS that consist of four CSs (CS1,C S2,
CS3and C S4). Each CS has a RAT table that shows the
current available resources with the CS, and the allocated
processes. For example, in CS1, the CS has four different
resources (R1,R2) that are assigned to ES1, and R3,R4that
are assigned to ES2. The resources processing time for each
resources is shown for each CS, For instance, in CS1, the
processing time needed for R2is equal to 100 ms, where it is
equal to 50 ms in CS2, which is expected since in C S2, each
ES is dedicated for one resource, which in turn will result in
less processing time. The example below shows that the SoS
is currently handling two processes, P0, and P1. The RAT
tables show how the resources are allocated between different
CSs. For example, as shown in the Constraints column of
RAT1, P0had the following constraints: an E2E delay equals
to 300 ms, and it needs the following resources in order: R1,
R2, and R5. Currently it is scheduled as follows:
It will be served first by CS1:R1, the busy time slot will
be from 0 to 75 ms, and the CS-route is as follows: CS1,
CS2, and then C S2again. One can notice that the needed
resources for this process have been reserved in the respected
CSs. For instance, after the process is served by CS1:R1, it
will be routed to CS2and be served by R2, the busy time
slot for CS2:R2will be from 75 to 125 ms. Notice that the
processing time for CS2:R2is equal to 50 ms according
to the RPT entry. Further, to simplify the scheduling prob-
lem, we ignored the communication and transmission delay
between CSs. Finally, the process will be delivered to CS2
again, where it will occupy R5from 125-175 ms. Here the
processing time for CS2:R5is equal to 50 ms.
Once a resource reservation request arrives, the admission
control manager at the receiving CS checks the resources
utilization trees for all the resources in all CSs, and assigns
6VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
Resources' discovery/ RAT
updates
The Selected
CS meets the
request resource's
constraints?
Process arrives using a Poisson
distribution arrival time
Exchange RAT with other CSs
Process a Pre-admission using
RRP
The process is admitted
Resources are released upon
process completion
Yes
No
Resources' CSs path
determination
The PRP sends RRR to the next
available CS that has the
requested resource
Reserve the resource on the
selected CS
All required
resources by the
process are
reserved?
Yes
No
Start reserving the process's
requested resources
FIGURE 4: The flow chart of the proposed admission control
algorithm.
the required resources to the available ones such that the
assignment process will meet the process constraints. In this
example, the admission control manager of CS1will attempt
first to assign P1to CS1, since C S1:R3is free, then it will
attempt to find the second needed resources (R1) to a suitable
CS, in this case, there is two available free resources for
R1, one in CS1with RTP equals to 75 ms, and another one
in CS4with RPT equals to 100 ms. The admission control
manager will chose the one with lowest RTP (i.e. CS3), after
that, the admission control manager will try to allocate the
last resource to P1(R4), which is allocated to CS2.
After discussing the RAT tables and entries. We will
discuss the process of admitting a new request (P2) which
assumed to arrive at the same time of the previous two
processes. The process P2request table is shown where the
process was originated from CS0, it is requesting the follow-
ing resources (R1,R2and R5), the process has a 250 ms
E2E delay constraint. Finally, the process priority is equal to
1. In order to decided whether the process can be admitted or
not, each CS will prepare a Resources Utilization Tree (RUT)
for all available resources in the SoSs. As shown in Fig. 7, a
sample RUT is constructed for R1. The RUT consists of a
parent node (the resource ID) and child nodes (CSs that have
the resource). On each child node, a Resource Utilization
Vector (RUV) in the form of a linked list is established. This
RUV has the following entries: the first entry is the RPT of
that resource in the respective CS. The second entry is a flag
indicating whether the resource is free (F) or busy (B), if it is
busy, then the reservation schedule as a function of the RPT is
shown. For example, in Fig. 6, R1is reserved in the first time
slot (from 0 to 75 ms, marked in red-color), while it is free
in the second and others time slots (marked in green color),
while R1in CS3is busy in the second time slot. Finally, R1
in CS4is free in all time slots. Note that the time slot duration
is a function of the RPT of the respective CS. Thus it is 75
ms in both CS1and C S3, while it is 100 ms in C S4.
Note that the assignment process may become more com-
plicated, especially if the goal is to perform the resources
allocation in an optimal manner. Another challenge appears
more than one request arrive at the same time with different
priorities. Then the admission process has to take into ac-
count not only the available resources, but also the process
priorities. In some case, it may issue a resource revocation
command to an assigned process to a specific resource with
lower priority, to allow higher priority processes to be served.
However, this revocation process should be done without
jeopardizing the constraints of the lower priority process. For
example, if a new request arrives (P3) as shown in Table 1,
with high priority (PL=1) and with a very strict E2E delay
(150 ms), which requires two resources (R1,R2). Then if
the admission control manager assigns the free available
resources that exists on CS4,C S1, respectively, then the
process will miss its E2E delay constraints, since R1and R2
in these CSs require 200 ms processing time, which is higher
than the 150 ms E2E delay constraints of P3. However,
according to the priority level of P3, the admission control
manager will give the priority to P3compared to the P0and
P1(priority level is 2 and 3, respectively). Then, the resources
requested by P3will have higher priority than P0and P1.
Thus, P3may get admitted and both P0and P1may get
executed later (if they will still meet their E2E delays) or
blocked.
VOLUME 4, 2016 7
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
Initiator CSM CS1 CSM
Process arrival
Resources' discovery
CS Resources' Allocation Table (RAT)
Resource Request Reservation (RRR)
RRR Response
Resource Reservation (if aplicable)
Resources Allocation Manager
Resources Allocation Manager
CS2 CSM CSn CSM
......
Resources Allocation Manager
Resources Allocation Manager
FIGURE 5: The UML sequence disgram of the proposed admission control algorithm.
ES1
(R1, R2) Router
CSM
Router
ES2
(R3, R4)
Constituent
System: CS 1
Router
CSM
Router
Constituent
System: CS 2
CSNS
ES4
(R3)
CSNS
Router
NMS
Router
MPLS Network Domain
MNS
Router
Router
Router
CSM
Router
Constituent
System: CS3
CSNS
Router
CSM
Router
Constituent
System: CS4
CSNS
RAT1
RID RPT RS PL Cons. CS-Route
R1 75 B:P0:BS
[0 -75]
2 E2E 350,
R1,R2,
R5
CS1, CS2, CS2
R2 100 F
R3 75 B:P1:BS
[0 -75]
3
E2E 250,R3, R1,
R4
CS1, CS3, CS2
R4 75 F
Process ID
SoS ID Resource ID(s) Priority level Constraints
P2 CS0 R1,R2, 1 E2E 150 ES1
(R1, R3)
ES2
(R4, R6)
ES2
(R1, R2)
ES1
(R5, R6)
ES3
(R2)
ES2
(R4)
ES1
(R5)
RAT2
RID RPT RS PL Cons. CS-Route
R2 50 B:P0:BS
[75 -125]
2 E2E 350,
R2,R5
CS2, CS2
R3 50 F
R4 50 B:P1:BS
[150 -200]
3 E2E 250
R4
CS2
R5 50 B:P0:BS
[125 - 175]
2 E2E 350,
R5
CS2
RAT4
RID RPT RS PL Cons. CS-Route
R1 100 F
R2 75 F
R5 75 F
R6 100 F
RAT3
RID
RPT RS PL Cons. CS-Route
R1 75 B:P1:BS
[75 -150]
3 E2E 250,
R1, R4
CS3, CS2
R3 100 F
R4 125 F
R6 75 F
FIGURE 6: Demonstration example of the admission control process and resources allocations.
TABLE 1: Process P3resources request.
Process ID Source SoS ID Resource ID(s) Priority level Constraints
P3CS0R1,R21 E2E delay 150
8VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
R1
CS1
RPT=75
B
[0-75]
[75- 150]
.
.
CS3
RPT=75
B
[75- 150]
[0-75]
.
.
CS4
RPT=100
F
[0-100]
.
.
[100- 200]
FIGURE 7: An example of the resource utilization tree of
resource R1.
V. SIMULATION AND PERFORMANCE EVALUATION
In this section, an SoS network architecture embedded with
the proposed admission control process and resources’ allo-
cation has been designed and evaluated. The SoS network
architecture has been simulated to mimic an SoS senario
with a realistic admission control and resources’ allocation
process. More specifically, the simulated SoS network ar-
chitecture comprises NCSs, each delivering some services
out of Spossible services within a certain period of time
based on the requirements of Pdifferent processes/requests.
Moreover, the simulated SoS network has been sufficiently
designed to address the dynamic nature of the requests and
the availability of the resources at any possible values of N,
S, and P.
For illustration purposes, an SoS network architecture with
N= 8 CSs has been considered; each can deliver a maxi-
mum of S= 6 services to fulfill P= 25 processes/requests
within the time horizon. It is worth mentioning that the
proposed admission control and resources’ allocation process
has been performed with a Matlab based simulator that has
been developed for this purpose. This simulator is capable
of simulating various SoS network architectures with any
possible random values of N,S, and P.
The generalizability of the designed SoS and its admission
control and resources’ allocation process entails the follow-
ing simulation aspects:
A. SERVICES PER CS
The number of services to be offered by each CS is
selected randomly from an arbitrary range that spans
the interval [5, 6], to ensure the complexity of the case
study being developed, where 5 and 6 are the minimum,
maximum number of services that a CS could deliver,
respectively.
The services to be offered by each CS are selected
uniformly at random, without replacement, from the
available S= 6 services.
Each service is assumed to be delivered by each CS
within a period of time that spans the interval [15, 50]
ms with a step size of 5 ms, where 15 and 50 are
the minimum, maximum time duration, respectively by
which a certain CS could deliver the service.
B. SERVICES PER PROCESS/REQUEST
The number of services to be requested by each pro-
cess/request is selected randomly from an arbitrary
range of values that span the interval [1, 6], where 1,6
are the minimum, maximum number of services that a
process could request, respectively.
The order of services to be requested by each process is
randomly initiated.
The priority level of each request is selected randomly
from an arbitrary range of values that span the inter-
val [1, 3], where 1 and 3 are the minimum and the
maximum priority level, respectively implied by each
process/request.
The E2E delay of each process/request is selected ran-
domly from an arbitrary range of values that span the
interval [300, 450] ms, with a step size of 10 ms, where
300 and 450 are the minimum, maximum E2E delay
constraint, respectively implied by each process/request.
The number of processes/requests to arrive at a time is
selected randomly from an arbitrary range of values that
span the interval [1, 2], where 1 and 2 are the minimum,
maximum number of processes/requests, respectively
that could be arrived at a time instant.
The processes/requests arrive according at Poisson dis-
tribution with a parameter (λ) equals to 20.
For clarification purposes, Table 2 and Table 3 show
the randomly generated CSs and processes/requests, respec-
tively. For instance, looking at Table 2, one can notice that the
first CS (CS1) can deliver the six considered services (i.e.,
R1R6), each with a particular processing time. For exam-
ple, the processing time needed for R1by CS1is equal to 25
ms, whereas the last CS (CS8) can only deliver five services
(i.e., all except R5), each with a particular processing time.
For example, the processing time needed for R1by CS8is
equal to 40 ms.
Furthermore, Table 3 reports the simulated random pro-
cesses/requests received at the SoS under consideration, their
requested constraints, i.e., execution order, priority, and the
E2E delay. For instance, the second service (P2) to be re-
ceived either individually or together with the previous or
the other subsequent process (e.g., P2or P2in this case,
respectively) requires the execution of the following services
in a chronological order as follows: R1,R3,R6, then R4.
It necessitates the execution of the above-mentioned services
with a maximum E2E equals to 300 ms. Last, the priority
level of P2is low (e.g., is equal to the lower bound 3), which
entails that receiving two processes at the same time (i.e., P2
VOLUME 4, 2016 9
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
and P3) implies the execution of either P3or P2since the
priority level of P3is high (i.e., 1).
The admission control process and resources’ allocation
continues until the execution (or blocking) of the consid-
ered simulated processes/requests (i.e., P= 25). The former
(execution) entails that it/there exist(s) CS/CSs that is/are
capable of handling the services required by the received
process(es) respecting the E2E delay constraint, whereas the
latter (blocking) entails that there are no CSs capable of
managing the required services while respecting the E2E
delay or there are no CSs capable of delivering the requested
services.
Once the whole processes/requests are received and man-
aged, the performance of the proposed admission control
process and resources’ allocation is evaluated. Fig. 8 shows
the executed processes/requests (depicted as bars in the Fig-
ure) and the blocked processes/requests (depicted as x in
the Figure). The order of the received processes/requests in
this illustrative example is shown on the x-axis. The Figure
also highlights the gain in time of those processes/requests
executed by utilizing admission control algorithm. The Fig-
ure also depicts: the average, minimum and maximum gain
in time which are illustrated by a dashed line, a yellow
highlighted bar and a red highlighted bar, respectively.
P1P3P2P4P5P7P6P8P9P10 P11 P12 P14 P13 P15 P16 P17 P1 8 P19 P21 P20 P22 P23 P24 P25
Processes
0
50
100
150
200
250
300
350
400
Gain in Time [ms]
Gain in Time per Process
Blocked Processes
Minimum Gain in Time
Maximum Gain in Time
Average Gain in Time
FIGURE 8: The gain in time and its characteristics of the
executed processes/requests as well as the blocked pro-
cesses/requests.
In summary, Table 4 reports the above-mentioned SoS
performance metrics used in the simulation which shows
a considerable average gain in time (for illustration, the
average gain in time is computed as per Eq. (1)) and a low
blocking probability (for illustration, the average gain in time
is computed as per Eq.(2)), thus illustrating the effectiveness
of the proposed admission control algorithm.
Gain in Time [ms] =E2E Actual Execution Time (1)
Blocking Probability [%] =Number of Blocked Processes
P100%
(2)
Fig. 9 shows the actual execution time and the E2E delays
of the executed processes/requests as depicted in Fig. 9a, and
the corresponding gain in execution time as depicted in Fig.
9b.
Looking at Fig. 9, one can recognize the following:
The time necessities to execute/fulfil the received pro-
cesses/requests increases with time due to the fact that
the resources offered by the available CSs become oc-
cupied with time. This is clear while looking at Fig. 9a,
i.e., the actual execution time approaches the E2E of the
executed/blocked processes/requests, and, consequently
at Fig. 9b, i.e., the gain in execution time decreases
as long as more processes/requests received and exe-
cuted/blocked with time.
The processes/request P24 is blocked due to the fact
the CSs available to fulfil the process’ services (i.e., six
services as per Table 3) require more time (i.e., 411 ms)
than the required E2E of the process/request P24 (i.e.,
400 ms). Thus, P24 is blocked.
Fig. 10 shows the dynamic nature of the received pro-
cesses/requests and the availability of the resources. The
Figure shows the P= 25 processes/requests, their requested
services (shown in different colors), and their admission
control process and resource allocation over the available
CSs (i.e., indicated as a number inside each service). The
blocked process(es)/request(s) is (are) shown in the Figure
as well. Furthermore, Fig. 10 provides an insights about the
dynamic nature of the received processes/requests and the
availability of the CSs’ resources as discussed below:
At t= 0, the first process/request (P1) is received.
It requires the execution of the following services in
order (Table 3): R1,R3,R6,R5with a priority level of
execution equals to 3 and E2E delay equals to 410 ms.
The CSs that fulfil the process/request’s services with
minimum execution time are occupied, they are: CS1,
CS2,C S3,C S3, respectively. It is worth to mention that
these CSs will be occupied for the requested services of
the P1until they have been fully fulfilled.
Two processes/requests have been received later at time
t= 14 ms and executed on time, i.e., P12 and P3.
However, since P3is with a high priority level (i.e., 1)
in execution with respect to the P2(i.e., 3), it will be ex-
ecuted first, as shown in the Figure. Notice, for instance,
that P3requires the execution of service R1first, thus the
CS that offers this service with the minimum execution
time is CS1and it will be now reserved for P3, while at
the same time P2also requires the service R1, but the
next best option for P2/R1is CS3.
The remaining processes/requests have continued to be
received and executed/blocked on timely basis follow-
ing a Poisson distribution function depending on the
CSs resources availability and their ability to fulfil the
processes/request E2E delay constrains.
Last, it is worth mentioning that no time slots have been
reserved for the blocked P24 process/request.
To further study the performance of the proposed admis-
sion control process and resource allocation, a sensitivity
10 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
TABLE 2: The simulated CSs and their available services and delivery time
Constituent Systems (CSs)
CS1C S2CS3)CS4C S5CS6CS7C S8)
Services
R125 40 30 NA 45 50 45 40
R230 45 20 35 40 40 20 50
R335 15 35 45 20 35 40 30
R415 20 40 25 25 25 15 35
R545 50 25 40 30 45 30 NA
R650 30 15 30 15 30 50 15
P1P3P2P4P5P7P6P8P9P10 P11 P12 P14 P1 3 P15 P16 P17 P18 P19 P2 1 P20 P22 P23 P24 P2 5
Processes
0
100
200
300
400
500
Excuation Time [ms]
Actual Execution Time [ms]
E2E [ms]
(a)
P1P3P2P4P5P7P6P8P9P10 P11 P12 P14 P1 3 P15 P16 P17 P18 P1 9 P21 P20 P22 P23 P2 4 P25
Processes
0
50
100
150
200
250
300
350
400
Gain in Time [ms]
Gain in Time per Process
Blocked Processes
(b)
FIGURE 9: The actual execution time and E2E of the whole simulated processes/requests (a) and their gain in time (b).
test has been carried out. Specifically, the influence of two
parameters of the SoS network architecture on the gain in the
execution time (in ms) and the blockage probability has been
investigated. The two parameters are:
1) The number of Constituent Systems (CSs). The pos-
sible number of CSs is assumed to cover the interval
[1,10], where the lower and upper bounds are the min-
imum and the maximum number of CSs that could be
available in the SoS network architecture, respectively.
2) The number of resources required by each received
processe(s)/request(s) (RPP). The possible number of
resources is assumed to cover the interval [1,6], where
the lower and upper bounds are the minimum and the
maximum number of resources that could be requested
by each received processe(s)/request(s), respectively.
To this aim, the admission control and resource allocation
process is simulated 100 times for each possible combination
of the two above-mentioned parameters, considering P= 10
processes/requests to be received adaptively with the time.
Once the 100 simulation trials are completed, the ultimate
gain in execution time and the blockage probability are cal-
culated by averaging their values across the 100 simulation
results.
Fig. 11 shows the gain in execution time (in ms) versus the
possible number of CSs available while varying the number
of resources required by each received process (s)/request(s),
RPP = 1 to RPP = 6. Looking at Fig. 11, one can recognize
the following:
For a particular RPP, as long as the number of available
CSs increases, the gain in execution time increases, as
expected.
For a particular number of CSs available, as long as
the number of resources required by each received pro-
cess(s)/request(s) (RPP) increases, the gain in execution
time decreases, as expected.
Similarly, Fig. 12 shows the influence of those two param-
eters on the blockage probability. Looking at Fig. 12, one can
notice the following:
For a particular RPP, as long as the number of CSs
available increases, the blockage probability decreases,
as expected.
For a particular number of CSs available, as long as
the number of resources required by each received pro-
cess(s)/request(s) (RPP) increases, the blockage proba-
bility increases, as expected.
VOLUME 4, 2016 11
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
FIGURE 10: The dynamic nature of the proposed admission control process and resource allocation of the received P= 25
processes/requests.
TABLE 3: The simulated processes/requests, execution or-
der, priority, and E2E
Requested services and Constraints
R1R2R3R4R5R6PL E2E
Processes/Requests
P10*120*4 3 3 410
P21 0*240*3 3 300
P31 0*320*0*1 450
P4130*0*2 4 3 420
P50*0*0*3 1 2 3 360
P63 0*120*0*3 400
P70*0*0*0*1 0*1 410
P8240*1 0*3 1 450
P96 1 2 5 4 3 2 360
P10 0*0*0*1 0*0*2 300
P11 320*1 4 0*1 390
P12 140*3 2 0*1 430
P13 0*1 0*0*2 0*3 330
P14 2310*4 5 2 410
P15 0*2310*0*1 380
P16 0*0*1 0*0*0*3 350
P17 6 3 2 1 4 5 3 340
P18 1 2 3 5 4 6 2 370
P19 120*0*0*0*3 360
P20 1 0*3 0*0*2 3 370
P21 0*1 0*3 4 2 1 360
P22 2 0*0*1 0*0*3 380
P23 3 2 4 5 1 6 3 420
P24 3 1 4 5 6 2 1 400
P25 1 0*0*0*0*0*1 450
*The service is not requested by a Process Pi, i = 1,...,P
TABLE 4: The proposed SoS performance metric
Minimum gain in time 28 ms
Maximum gain in time 381 ms
Average gain in time 175.958 ms
Blocking probability 4%
CS1CS2CS 3CS4CS5CS 6CS7CS8CS 9CS10
Number of available CSs
140
160
180
200
220
240
260
280
300
320
Gain in Time [ms]
RPP=1
RPP=2
RPP=3
RPP=4
RPP=5
RPP=6
FIGURE 11: The influence of the number of CSs available
and the number of resources required by each received pro-
cesse(s)/request(s) on the gain in execution time.
VI. CONCLUSIONS AND FUTURE WORK
This work studied the admission control and resources’ allo-
cation process in SoS paradigm, which comprise various sys-
12 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
CS1CS2CS 3CS4CS5CS 6CS7CS 8CS9CS10
Number of available CSs
0
10
20
30
40
50
60
70
80
90
Blocking Probability [%]
RPP=1
RPP=2
RPP=3
RPP=4
RPP=5
RPP=6
FIGURE 12: The influence of the number of CSs available
and the number of resources required by each received pro-
cess(s)/request(s) on the blockage probability.
tems (constitute systems) working independently from each
other to fulfill any complex task received. The coordination
between these systems is crucial to fulfilling the received task
requirements and constraints. To this aim, this work proposes
an admission control and resources’ allocation process to
effectively manage any complex task received by distributing
it among the proper working systems that can fulfill its
constraints with fewer efforts. The proposed approach has
been verified concerning a simulated case study that was
appropriately developed to mimic a realistic SoS paradigm
embedded with realistic admission control and resources’
allocation. The performance of the proposed algorithm has
been investigated in terms of gain in the execution time
and blockage probability. The former entails highlighting the
time gained in executing the received task by the available
systems of the SoS, whereas the latter entails stating the
services that have been requested but blocked due to various
reasons, like, for example, the non-availability of the systems
that could fulfill the requested services. Further, a sensitivity
analysis has been carried out to study the effect of having dif-
ferent possible numbers of available systems and the number
of services requested on the overall efficacy of the proposed
process. As a future work, we are investigating more complex
cases where requests may have multiple constraints that
should be fulfilled jointly while maximizing the resources’
utilization and minimizing the blockage probability.
REFERENCES
[1] F. C. Delicato, P. F. Pires, T. Batista, E. Cavalcante,B. Costa, and T. Barros,
“Towards an iot ecosystem,” in Proceedings of the First International
Workshop on Software Engineering for Systems-of-Systems, 2013, pp.
25–28.
[2] R. M. Tanash, A. K. Khalifeh, and K. A. Darabkh, “Communication
over cloud computing: A security survey,” in 2019 42nd International
Convention on Information and Communication Technology, Electronics
and Microelectronics (MIPRO). IEEE, 2019, pp. 496–501.
[3] K. A. Kiblawi, Tarek and K. A. Darabkh, “Disruptive innovations in
cloud computing and their impact on business and technology, in 2015
4th International Conference on Reliability, Infocom Technologies and
Optimization (ICRITO)(Trends and Future Directions). IEEE, 2015, pp.
1–4.
[4] B. K. Tannahill and M. Jamshidi, “System of systems and big data
analytics–bridging the gap,” Computers & Electrical Engineering, vol. 40,
no. 1, pp. 2–15, 2014.
[5] E. Alonso, N. Karcanias, and A. G. Hessami, “Multi-agent systems: A new
paradigm for systems of systems,” 2013.
[6] A. Khalifeh, R. Obermaisser, D. e. D. I. Abou-Tair, and M. Abuteir,
“Systems-of-systems framework for providing real-time patient monitor-
ing and care,” in Proceedings of the 8th International Conference on
Pervasive Computing Technologies for Healthcare, 2014, pp. 426–429.
[7] A. Khalifeh, Z. Sakkijha, H. Qaderi, O. Ghatasheh, A. Issa, A. Shatat,
H. Jaber, R. Alwardat, S. Alhaj-Ali et al., An energy efficient wsn
implementation for monitoring and critical event detection, in 2019
2nd IEEE Middle East and North Africa COMMunications Conference
(MENACOMM). IEEE, 2019, pp. 1–6.
[8] J. Mahalakshmi, K. Kuppusamy, C. Kaleeswari, and P. Maheswari, “Iot
sensor-based smart agricultural system,” in Emerging Technologies for
Agriculture and Environment. Springer, 2020, pp. 39–52.
[9] H. Liu, Y. Tian, Y. Gao, J. Bai, and J. Zheng, “System of systems oriented
flight vehicle conceptual design: Perspectives and progresses, Chinese
Journal of Aeronautics, vol. 28, no. 3, pp. 617–635, 2015.
[10] A. W. Colombo, T. Bangemann, and S. Karnouskos, “A system of systems
view on collaborative industrial automation, in 2013 IEEE international
conference on industrial technology (ICIT). IEEE, 2013, pp. 1968–1975.
[11] M. W. Maier, Architecting principles for systems-of-systems,” Systems
Engineering, vol. 1, no. 4, pp. 267–284, 1998.
[12] A. Bondavalli, S. Bouchenak, and H. Kopetz, Cyber-Physical Systems of
Systems, Foundations A Conceptual Model and Some Derivations: The
AMADEOS Legacy. Springer, 2016.
[13] M. Jamshidi, Systems of Systems Engineering—Innovations for the 21st
Century. Wiley, 2009.
[14] R. Obermaisser, M. Abuteir, A. Khalifeh, and I. Dhiah el Diehn, “Systems-
of-systems framework for providing real-time patient monitoring and care:
challenges and solutions,” in ICTs for Improving Patients Rehabilitation
Research Techniques. Springer, 2014, pp. 129–142.
[15] R. Obermaisser, M. Abuteir, A. Khalifeh, and D. e. D. I. Abou-
Tair, Systems-of-Systems Framework for Providing Real-Time Patient
Monitoring and Care: Challenges and Solutions. Berlin, Heidelberg:
Springer Berlin Heidelberg, 2015, pp. 129–142. [Online]. Available:
https://doi.org/10.1007/978-3-662-48645- 0_12
[16] A. Khalifeh, R. Obermaisser, D. e. D. I. Abou-Tair, and
M. Abuteir, “Systems-of-systems framework for providing real-
time patient monitoring and care,” in Proceedings of the 8th
International Conference on Pervasive Computing Technologies for
Healthcare, ser. PervasiveHealth ’14. ICST, Brussels, Belgium,
Belgium: ICST (Institute for Computer Sciences, Social-Informatics
and Telecommunications Engineering), 2014, pp. 426–429. [Online].
Available: http://dx.doi.org/10.4108/icst.pervasivehealth.2014.255324
[17] D. Abou-Tair, S. Alouneh, A. Khalifeh, and R. Obermaisser, “A security
framework for systems-of-systems, in Advances in Computer Science
and Ubiquitous Computing, J. J. Park, V. Loia, G. Yi, and Y. Sung, Eds.
Singapore: Springer Singapore, 2018, pp. 427–432.
[18] S. Alouneh, D. e. D. I. Abou-Tair, A. Khalifeh, and R. Obermaisser,
“Service identification framework for systems of systems based on mpls
technology, in Advances in Computer Science and Ubiquitous Comput-
ing, J. J. Park, V. Loia, G. Yi, and Y. Sung, Eds. Singapore: Springer
Singapore, 2018, pp. 359–364.
[19] D. e. D. I. Abou-Tair, A. Khalifeh, S. Alouneh, and R. Obermaisser,
“Incremental, distributed, and concurrent service coordination for reliable
and deterministic systems-of-systems,” IEEE Systems Journal, vol. 15,
no. 2, pp. 2470–2481, 2021.
[20] A. Khalifeh, D. Abou-Tair, S. Alouneh, R. Obermaisser, and M. Schmidt,
“Admission control and resource allocation for distributed services in
system-of-systems: Challenges and potential solutions,” in 2018 IEEE
13th International Symposium on Industrial Embedded Systems (SIES).
IEEE, 2018, pp. 1–4.
[21] S. Sharma and P. Kuila, “Design of dependable task scheduling
algorithm in cloud environment, in Proceedings of the Third International
Symposium on Women in Computing and Informatics, ser. WCI ’15.
New York, NY, USA: ACM, 2015, pp. 516–521. [Online]. Available:
http://doi.acm.org/10.1145/2791405.2791502
VOLUME 4, 2016 13
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
[22] J.-J. Hwang, Y.-C. Chow, F. D. Anger, and C.-Y. Lee, “Scheduling
precedence graphs in systems with interprocessor communication times,”
SIAM J. Comput., vol. 18, no. 2, pp. 244–257, Apr. 1989. [Online].
Available: http://dx.doi.org/10.1137/0218016
[23] G. C. Sih and E. A. Lee, “A compile-time scheduling heuristic for
interconnection-constrained heterogeneous processor architectures,” IEEE
Trans. Parallel Distrib. Syst., vol. 4, no. 2, pp. 175–187, Feb. 1993.
[Online]. Available: http://dx.doi.org/10.1109/71.207593
[24] D. Warneke and O. Kao, “Exploiting dynamic resource allocation for
efficient parallel data processing in the cloud,” IEEE Trans. Parallel
Distrib. Syst., vol. 22, no. 6, pp. 985–997, Jun. 2011. [Online]. Available:
http://dx.doi.org/10.1109/TPDS.2011.65
[25] S. Singh and I. Chana, “A survey on resource scheduling in cloud
computing: Issues and challenges,” Journal of Grid Computing, vol. 14,
no. 2, pp. 217–264, 2016. [Online]. Available: http://dx.doi.org/10.1007/
s10723-015- 9359-2
[26] Y. Huang, Y. Wang, and N. Liu, “Low-carbon economic dispatch and
energy sharing method of multiple integrated energy systems from
the perspective of system of systems, Energy, vol. 244, p. 122717,
2022. [Online]. Available: https://www.sciencedirect.com/science/article/
pii/S0360544221029662
[27] P. Shiva Prakasha, N. Naeem, P. Ratei, and B. Nagel, “Aircraft architecture
and fleet assessment framework for urban air mobility using a system
of systems approach,” Aerospace Science and Technology, p. 107072,
2021. [Online]. Available: https://www.sciencedirect.com/science/article/
pii/S1270963821005824
[28] M. Mennenga, L. Büth, F. Cerdas, and C. Herrmann, “Synthetic
emergence as a functional unit for the environmental assessment of a
system of systems,” Procedia CIRP, vol. 90, pp. 393–398, 2020, 27th
CIRP Life Cycle Engineering Conference (LCE2020)Advancing Life
Cycle Engineering : from technological eco-efficiency to technology
that supports a world that meets the development goals and the
absolute sustainability. [Online]. Available: https://www.sciencedirect.
com/science/article/pii/S2212827120301451
[29] S. Majidi and R. Obermaisser, “Evolutionary algorithm for incremental
scheduling in systems of systems with real-time requirements,” in IECON
2021 47th Annual Conference of the IEEE Industrial Electronics Society,
2021, pp. 1–6.
[30] S. K. Roy, R. Devaraj, A. Sarkar, K. Maji, and S. Sinha, “Contention-
aware optimal scheduling of real-time precedence-constrained task
graphs on heterogeneous distributed systems,” Journal of Systems
Architecture, vol. 105, p. 101706, 2020. [Online]. Available: https:
//www.sciencedirect.com/science/article/pii/S1383762119305132
[31] S. K. Roy, R. Devaraj, A. Sarkar, and D. Senapati, “Slaqa: Quality-level
aware scheduling of task graphs on heterogeneous distributed systems,
ACM Trans. Embed. Comput. Syst., vol. 20, no. 5, jul 2021. [Online].
Available: https://doi.org/10.1145/3462776
[32] S. K. Roy, R. Devaraj, and A. Sarkar, “Contention cognizant scheduling of
task graphs on shared bus-based heterogeneous platforms,” IEEE Trans-
actions on Computer-Aided Design of Integrated Circuits and Systems,
vol. 41, no. 2, pp. 281–293, 2022.
[33] C. Guariniello, T. B. Marsh, R. Porter, C. Crumbly, and D. A. DeLaurentis,
“Artificial intelligence agents to support data mining for sos modeling of
space systems design,” in 2020 IEEE Aerospace Conference, 2020, pp.
1–11.
[34] Z. Fang, J. Liao, and X. Zhou, “Improving system-of-systems agility
through dynamic reconfiguration,” in 2020 IEEE International Conference
on Systems, Man, and Cybernetics (SMC), 2020, pp. 4466–4472.
[35] P. Mazzetti, M. A. Latre, M. Bauer, R. Brumana, S. Braumann, and
S. Nativi, “Energic-od virtual hubs: A brokered architecture for facilitating
open data sharing and use,” in Proc. of eChallenges e-2015 Conference,
2015, pp. 1–11.
DHIAH EL DIEHN I. ABOU-TAIR is an asso-
ciate professor at the German-Jordanian Univer-
sity (GJU), School of Electrical Engineering and
Information Technology, received his PhD from
the University of Siegen, Germany. Dr. Abou-
Tair research focus is on privacy laws and regula-
tions adoption in information systems through an
ontology-based approach. Currently, his research
concentrates on the areas of privacy-enhancing
technologies, security, System of Systems, and
Internet of Things. Dr. Abou-Tair has been involved in several EU and
German-funded research and capacity building projects.
ALA’ KHALIFEH received the PhD degree in
Electrical and Computer Engineering from the
University of California, Irvine -USA in 2010. He
is currently an associate professor in the Commu-
nication Engineering department at the German
Jordanian University. He is currently the IEEE
Jordan section chair. His research is in communi-
cations technology, and networking with particu-
lar emphasis on optimal resource allocations for
multimedia transmission over wired and wireless
networks, Internet of Things and wireless sensor networks.
SAMEER AL-DAHIDI received the B.Sc. degree
(with honors 1st rank) in Electrical and Com-
puter Engineering from The Hashemite Univer-
sity, Zarqa, Jordan, in 2008, the M.Sc. degree
(very good 1st rank) in Nuclear Energy (Oper-
ations Specialty) from Ecole Centrale Paris and
Université Paris- Sud 11, Paris, France in 2012,
and the Ph.D. degree (with honors) in Energy and
Nuclear Science and Technology from Politecnico
di Milano, Milan, Italy in 2016. In 2008- 2010,
he worked as an Electrical Instruments Engineer at CCIC and Kharafi
National in Oil Gas and petrochemical mega projects in Kuwait and UAE.
He is currently an Associate Professor at the Mechanical and Mainte-
nance Engineering Department, School of Applied Technical Sciences at
the German Jordanian University, Amman, Jordan. His current research
interests include the development of analytics and models for Prognostics
and Health Management (PHM), operation, maintenance and Reliability,
Availability, Maintainability, and Safety (RAMS) analysis of engineering
systems, and the development of Artificial Intelligence (AI)-based methods
for renewable energy production prediction. Besides, Dr. Al-Dahidi has
interests in renewable energy systems and mechanical Ventilation and Air-
Conditioning (HVAC), and others. Dr. Al- Dahidi is the author and co-
author of more than 40 papers on high-quality international journals and
proceedings of international conferences, a referee of more than 20, and a
guest editor of 3 special issues at high-quality international journals. Dr. Al-
Dahidi supervised and co-supervised more than 15 Bachelor, Master, and
Ph.D. thesis and dissertations in Jordan and Italy.
14 VOLUME 4, 2016
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
Abou-Tair et al.: Coordination Protocol and Admission Control for Distributed Services in System-of-Systems
SAHEL ALOUNEH is full professor of computer
engineering at Al Ain University. He held the post
of Dean of the Faculty of electrical engineering
and information technology and the Dean of Sci-
entific Research at the German Jordanian Uni-
versity. His research interests include computer
and communication networks, big data security,
cloud computing, software security, MPLS secu-
rity and recovery, wireless networks security, soft-
ware testing, computer design, and architecture.
ROMAN OBERMAISSER is full professor at
the Division for Embedded Systems of University
of Siegen. He has studied computer sciences at
Vienna University of Technology, and received
the Master’s degree in 2001. In 2004, Roman
Obermaisser has finished his doctoral studies in
Computer Science with Prof. Hermann Kopetz
at Vienna University of Technology as research
advisor. In 2009, Roman Obermaisser has received
the habilitation ("Venia docendi") certificate for
Technical Computer Science. His research work focuses on system archi-
tectures for distributed embedded real-time systems. He wrote a book on an
integrated time-triggered architecture published by Springer-Verlag, USA.
He is the author of several journal papers and conference publications. He
has also participated in numerous EU research projects (e.g., SAFEPOWER,
universAAL, DECOS, NextTTA) and was the coordinator of the European
research projects DREAMS, GENESYS and ACROSS.
VOLUME 4, 2016 15
This article has been accepted for publication in IEEE Access. This is the author's version which has not been fully edited and
content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2022.3207550
This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see https://creativecommons.org/licenses/by/4.0/
ResearchGate has not been able to resolve any citations for this publication.
Article
Full-text available
This research article explores Urban Air Mobility (UAM) from a System of Systems (SoS) perspective in order to understand the impact of different fully electric UAM aircraft architectures on the overall SoS capability. For this purpose, a framework, combining aircraft design methods with an agent-based simulation, is developed. Thereby, not only different UAM aircraft architectures, but also fleet combinations, technology scenarios, and operational strategies are studied and evaluated for different success criteria. The UAM fleets are simulated for 24-hour operations, considering non-uniform passenger demand, dispatch of passenger as well as deadhead flights, aircraft architectural performance, load factor, energy consumption, and turnaround procedures. A large design of experiments, consisting of approximately 5,000 design points, is executed. Eventually, this article demonstrates the proof of concept for the proposed SoS framework and provides several parameter sensitivities for a given UAM scenario. For such complex SoS, analytical methods would not suffice for understanding complex and often nonlinear interactions. Therefore, the proposed simulation driven framework proves to be successful by providing sensitivity study results, linking subsystem, system (aircraft) and system of system (fleet) level. Thus, the framework allows for comprehensive understanding of the SoS design space and is important for successful deployment or optimization of UAM aircraft & fleet for a given city and operational context.
Article
Full-text available
Technical systems are increasingly complex and interconnected, forming higher level System of Systems (SoS). SoS provide emergent functions which are greater or different than the sum of the functions of their constituent systems. To be able to assess the environmental impact of an SoS, a functional unit is required. Therefore, we discuss the opportunities of synthetic emergence to serve as functional unit of SoS. A deep understanding of emergent properties and the function of SoS is given. Finally, a comprehensive discussion uses the cases of smart energy grids and urban symbiotic food production systems for initial reflections about the concept.
Article
This paper mainly focuses on the low-carbon economic dispatch and energy sharing method of multiple Integrated Energy Systems (IES) from the perspective of System of Systems. First, a new model of carbon quota allocation and trading considering carbon balance and carbon sharing among multiple systems is established, and a low-carbon economic dispatch and energy sharing framework based on efficient utilization of regional carbon quota is proposed. Then, a Stackelberg game-based optimization model is proposed for energy service provider (ESP) and IES groups based on the collaborative optimization of IES and carbon transaction cost, where the ESP acts as a leader who chooses proper interactive price to promote energy sharing within the system while IES serves as a follower who minimizes its operation cost in response to the interactive price set by ESP. Moreover, the existence and uniqueness of the Stackelberg equilibrium are analyzed and a decentralized solution algorithm is suggested to reach the goal of optimal energy sharing and optimal operation. Finally, the case study of a joint energy sharing system containing three regional IESs are provided to show the effectiveness of the proposed method. The results demonstrate that the proposed method can not only increase profit of ESP and IES, but helpful to promote the optimal utilization of energy resources and the optimal distribution of carbon emission quota among energy systems.
Article
Continuous demands for higher performance and reliability within stringent resource budgets is driving a shift from homogeneous to heterogeneous processing platforms for the implementation of today’s cyber-physical systems (CPSs). These CPSs are typically represented as Directed-acyclic Task Graph (DTG) due to the complex interactions between their functional components that are often distributed in nature. In this article, we consider the problem of scheduling a real-time application modelled as a single DTG, where tasks may have multiple implementations designated as quality-levels, with higher quality-levels producing more accurate results and contributing to higher rewards/Quality-of-Service for the system. First, we introduce an optimal solution using Integer Linear Programming (ILP) for a DTG with multiple quality-levels, to be executed on a heterogeneous distributed platform . However, this ILP-based optimal solution exhibits high computational complexity and does not scale for moderately large problem sizes. Hence, we propose two low-overhead heuristic algorithms called Global Slack Aware Quality-level Allocator ( G-SLAQA ) and Total Slack Aware Quality-level Allocator ( T-SLAQA ), which are able to produce satisfactorily efficient as well as fast solutions within a reasonable time. G-SLAQA , the baseline heuristic, is greedier and faster than its counter-part T-SLAQA , whose performance is at least as efficient as G-SLAQA . The efficiency of all the proposed schemes have been extensively evaluated through simulation-based experiments using benchmark and randomly generated DTGs. Through the case study of a real-world automotive traction controller , we generate schedules using our proposed schemes to demonstrate their practical applicability.
Article
Demands for high performance as well as reliability within stringent resource budgets is driving a shift from homogeneous to heterogeneous processing platforms for the implementation of today’s cyber-physical systems (CPSs). These CPSs are typically represented as Directed-acyclic Task Graphs (DTGs) due to the complex interactions between their functional components which are often distributed in nature. This work deals with the problem of scheduling a cyber-physical system modelled as DTG. First, we present an optimal solution using Integer Linear Programming (ILP) for the DTGs, to be executed on distributed heterogeneous processors which are interconnected via shared buses. However, this ILP based optimal solution exhibits high computational complexity and doesn’t scale for moderately large problem sizes. Hence, we propose a low-overhead heuristic algorithm called Contention Cognizant Task and Message Scheduler (CC-TMS), which is able to produce satisfactorily efficient as well as fast solutions within reasonable time. The efficiency of the proposed scheme has been extensively evaluated through simulation based experiments using benchmark DTGs. Through the case study of a real-world automotive traction controller, we demonstrate the practical applicability of our proposed scheme.
Article
Systems-of-systems (SoS) are gaining increasing at- tention for the realization of safety-relevant applications with re- liability and real-time requirements, by coordinating autonomous constituent systems from different application areas. For a given application that is initiated at a constituent system, the provision and use of services between constituent systems must be coor- dinated. This involves identifying constituent systems that can provide the services, optimizing the use of the services driven by extra-functional properties, performing admission control for service provision, reserving resources for provided services, and recursively using services from subcontracted constituent systems to realize service provisions. In this article, new conceptual models with service provisioning for reliable and deterministic SoS applica- tions, as well as, distributed algorithms for the discovery, optimized selection, and admission of service provisions are introduced. The proposed models will result in a two-step admission control algo- rithm with short- and long-term admissions based on schedulabiliy tests, resulting in service provisions between constituent systems. Furthermore, this article will investigate the integration of cloud resources for the on-demand instantiation of services. The models and the service coordination are evaluated in a use case from the health-care domain, which investigates the execution time and the scalability of the service coordination.