Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)

IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

3 –PHASE LEADER ELECTION

ALGORITHM FOR DISTRIBUTED SYSTEMS
Aparna Rajesh Atmakuri SivaSankarRao.S
Pushya Chaparala
Computer Science and Engineering Computer Science and Engineering
Computer Science and Engineering
Visveswaraya Technological University Guru Nank Institutions
Guru Nank Institutions
Karnataka, India Hyderabad, India
Hyderabad, India
[email protected] [email protected]
[email protected]

Abstract—Election plays a foremost part in distributed 2) Liveness Condition based on that every node will
systems. The reason behind the Election process in distributed eventually enter an elected or non- elected state.
system is to optimize in terms of load and improving the
efficiency of the system there by improving the process of
group communication. Choosing a node to be a leader in a
distributed environment is intricate. Over years, there are so
much of exploitations in the existing techniques such as ring
and bully but these algorithms could not able to provide an
appropriate solution in terms of message complexity. In this
paper, we proposed a novel method known as “3-Phase Leader
Election Algorithm” that consists of three phases. Firstly, it
identifies and filters the nodes for election. Secondly, the
filtered nodes are been validated for determination of prime
node via which the group communication takes place. Finally,
the Prime node will be identified and accepted in the
acceptance phase. Our approach is based on understanding the
complexity involved behind a message and improving the
efficiency of the system via following performance metrics such
as Time and Communication complexity.

Keywords—Ring, Bully, 3-PLE, Distributed Systems

Fig. 1 Election

I. INTRODUCTION Motivated by the above-mentioned assumptions and aspects,

Election is one of the important aspects of discussion in a in this paper we try to develop a novel approach for electing
distributed system. The idea behind the use of this side in a a node in the distributed network for communication
distributed system is to improve the efficiency and efficiency. Our approach named 3-Phase Leader election
algorithm:
effectiveness of group communication thereby minimizing
3PEL which aims in reducing the complexity involved in
the overhead involved in the information interchange
message communication from individual process. The
process. The principle behind this technique is to identifying algorithm is composed in three phases 1) election phase 2)
a Controlling node or Central coordinator from a given set voting phase 3) acceptance phase. The detail about each
of processes. The choosing of a prime node is based on phase has been discussed in rest of the paper.
some unique aspects such as priority Id, Load on the node, Section -2 of the paper is about literature review and its
Configuration parameters etc. Over years many researchers related work. Section-3 discusses the existing algorithms
started enhancing and exploiting the existing election such ring and bully. Section -4 presents our novel approach.
algorithms such as bully based and Ring based, but none of Section -5 showcases the analysis and the performance of
them could able to resolve the issue involved in the the proposed approach for leader election.
complexity of message involved during communication II. RELATED WORK
interchange. Another basic reason for electing a prime node
First Electing a leader as a part of group communication
is to construct a fault-tolerant distributed system. In general has been widely proposed and promoted over the last few
the each of the election algorithm devised so far differs in years. However, the postulate behind these election
terms of following four aspects first, based on arrangement algorithms is that all nodes required to be participate in the
of nodes in the network some of which are based on ring, leader election. Further, the participation of these nodes has
fully connected and spanning tree etc,. Second is the process lead to increase in the overhead of transmitting messages
with each other in anticipation of finding the leader for the
names, third is its Communication mechanism and finally
network [1-5]. Once an assessment of node to be elected as a
Size of the network. Apart from these each of the existing leader is made, all the other participants of the group will
algorithms is based on two postulates: accept the chosen node as a leader. The algorithms has been
1) Safety Condition: as is based on a solitaire node widely used and promoted in almost all types of distributed
can enter into the elected state and eventually become the system that ranges from IOT, WSN, MANET, VANET,
leader of the system. Cloud Computing etc., but each of these systems has used

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 898

 Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

the role of electing a leader in their own way that is suitable A. Bully Algorithm
for their information interchange. Paper [6], proposes a novel Garcia Molina in 1982 designed an election-based
approach of finding the leader using MAX-HEAPIFY () algorithm [1] and named it as bully. This algorithm uses a
procedure a part of heap tree method has been used. In [7], a Process ID for electing leader dynamically in a distributed
protocol that elects the local leader form subset or partition, computing environment. A leader process is elected based on
but it could able to solve the problem of communication the highest process ID of the system.
failures and also failed in electing a unique leader. In [8], it
maintains a global database about the necessary conditions of 1) Assumptions:
nodes to be a leader with an assumption the nodes uses a i. Every process of the system knows about the
reliable and asynchronous communication link. The authors other process “process ID” and is unique with
of this paper [9], tried to answer the problem by choosing a respect to each other.
prime node for a set of nodes with unique identifiers in a
ii. Processes are unaware of the status of the
fixed network within a reliable channel of communication. In
paper [10], the authors proposed an algorithm for election other processes.
based on candidate list in which they tried to reduce the iii. Process failure is based on timeouts and is
message passing in the process of election. But the algorithm synchronous with each other.
does not provide any clue of finding the next leader. In [11], iv. There may be chances of process failure even
the algorithm uses a system identification number (ID) to during execution of the algorithm.
order the list of candidates for election, and the leader is v. Time bound and reliable message delivery
elected from the available list. However the process does not
between processes.
provide any information regarding list maintenance cost.
Paper [12] is an enhancement of bully algorithm with fault
detection. This method detects crashed nodes at regular 2) Working Procedures:
intervals so that it can maintain current list of live nodes, but The objective of the bully algorithm is choosing a
it increased the message passing cost. A Consensus-Based process as a coordinator with an approval of other process of
Leader Election Algorithm [13] and Hierarchical Leader the system. On the other round, it is about electing the
Election Protocol Algorithm [14] that are been designed for process with highest Process-ID or highest priority. In case,
mobile ad hoc networks, where the former elects the leader Coordinator of the system fails, then the processes will
from preordered list however the algorithm does maintain the
immediately hold an election for choosing the leader. The
cost of the list and uses system ID for ordered list. The later
algorithm works in electing the leader by constructing a procedure is as follows:
backbone architecture on a clustered manet. Step1: Any process P of the system can initiate an election
procedure if it finds leader failure.
The ideology of this paper is to classify cluster heads into
backbone or leaf nodes through constructing a minimum i. This process P send an election message to those
spanning tree. The algorithm has two modes of operation process which has higher Process-IDs
hop-based and position based. There were also algorithms ii. If an acknowledgement or response is not reached
based on synchronous and asynchronous known as SEFA the initiated process P within the time bound then,
and AEFA [15] for electing leader in wireless ad hoc this process P will become the coordinator process
networks, SEFA based on security constraint and works in Step2: When a high priority process receives an election
hop-by-hop manner by choosing leader in each consecutive
hop by using a common evaluation function until it finalizes message from the initiated process P then:
the leader of the network. On the other hand, AEFA is a i. The high priority process sends an
non-secure election algorithm where the leader gets elected acknowledgement to the initiated process that it
based on the metric known as node weight value (NWV). will take over.
The node with highest NWV will be chosen as the leader. ii. Now, this high priority process holds an election to
However, these algorithms involve a large increase in its next higher process.
message and time complexity. There are also papers of
iii. The procedure will continue until it reaches the
electing leader in dynamic networks with lossy links [16-18].
Article [19], proposed a self-stabilizing leader election highest Priority-ID of the system.
algorithm with an approach of directed acyclic graph using iv. Finally, the new leader will announce itself as
TORA mechanism where the nodes will have partial coordinator of the system.
knowledge about the system and is mostly suitable for Step3: In case any other process which has higher priority-
MANETs. In [20], for anonymous and dynamic networks the ID in comparison with coordinator then again, the algorithm
method of choosing leader is based on time-varying runs.
communication graph with an assumption that network
remains connected over time. 3) Advantages:
 The probability of detecting a failed process in
III. EXISTING PROTOCOLS Bully algorithm is lowered in contrast to other
algorithm which leads in lowering of traffic loads.
Several algorithms are being designed for electing leader
in distributed systems. In this section we describe the  The process with high priority number can able to
classical election-based algorithms their assumptions of each detect the failure of the coordinator ad hence can
algorithm, working procedure, Advantages and be involved in the process of election.
disadvantages.

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 899

 Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

4) Limitations: IV. PROPOSED METHODOLOGY

 The algorithm involves five stages to detect crash In proposed method we are going to reduce the
in the process communication complexity that involves both time and also
 Massive message broadcasting during the election messages. The discussion of our methodology is based on
process three phases Election phase, Voting Phase and Acceptance
Phase.
B. Ring Algorithm
Ring Algorithm is a diversification of Bully algorithm. It A. Assumptions
is a logical ring with processes assigned with a position. a. Not more than 2 nodes will have the same system
Every process has an idea about it neighbor/ adjacent priority numbers.
process in the ring. The highest process Id in the ring is the b. Each node can have a system id, Priority id and its
coordinator of the ring. own configuration, workload.
1) Assumptions: c. A node can vote to show its support to the other
i. Between every two processes there must be nodes
one message hop. d. No node fails in the system.
ii. The arrangement of processes in the ring must e. Here the election begins in the new system not
be unidirectional when the coordinator fails.
iii. The network must be reliable f. Each node knows the information of all the nodes
present in the system in terms of their id’s.
2) Working Procedures g. A node with highest voting count will stand as a
Token Ring algorithm works: coordinator.
Step-1: Arrange the process in the form of ring each h. Her each process can broadcast the messages.
having its own unique ID
B. Used Messages and Variables
Step-2: Message passing must be unidirectional.
Step-3: When a Process X identifies the crashed
TABLE I. USED MESSGAES AND VARIABLES
coordinator, it creates an election message with its own
ID number Message Type Variable Functions
CRQ Sys-id, Filtered nodes will
Step-4: Process X sends the message to its immediate (Coordinator pri-id, voting-count, send this message to
live process Request Packet) visitors-id all the nodes in the
Step-5: As the message gets circulated in the ring, every system to inform their
priority and ask for
process will append its own Process-ID support( voting)
Step-6: When the message reaches to its initiated process CAM Sys-id, voting-count To declare itself as a
(Coordinator coordinator to
P, it identifies the highest Id in the message and the Acceptance everyone in the system
process with highest Id becomes the coordinator of the message)
ring.
C. Procedure
3) Advantages: Phase-1: Election Phase
 Message passing is much easier here because every “Nodes are filtered in this phase from the given list
of rules”
process in the ring knows the details of the process
Whenever the system is initialized each node will
to which it has to pass the message token. have its own configuration, load, sys-id, priority-id. This
 Since the token circulates among processes in a phase is considered as the first phase for our algorithm
well-defined order, starvation cannot occur. because here we are going to filter the nodes among the
nodes in the system. So that instead of every node
4) Limitations: broadcasting the CRQ message only the filtered nodes
 Message passing in the token ring is time will perform the broadcasting through this we can reduce
consuming. the Communication complexity.
 If the token is ever lost, it must be regenerated. In The rules that are listed for filtering the nodes are:
fact, detecting that it is lost is difficult, since the ER1: Node with highest configuration and highest
load are selected in the first priority.
amount of time between successive appearances of ER2: Node with highest configuration and less
the token on the network is not a constant. The fact load can also be given the priority
that the token has not been spotted for an hour does ER3: Nodes based on their performance- Job
not mean that it has been lost; some process may finishing are also given priority.
still be using it. With the help of above three election rules (ER) in the
first phase, almost 75% of the nodes will be filtered and
these filtered nodes are allowed to carry on the second
phase, where they broadcast the CRQ message to every

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 900

 Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

node in the system in order to gain the votes to declare 5. it will be announced as winner for coordinator.
themselves as coordinator. 6. Else if
7. Support received is also same then
8. Supported nodes priority will be summed and
the group with highest summed value nominator
will be declared as winner.
9. Else
10. Two nodes have the same priority id then
11. Voting count will be taken into consideration
and winner will be declared
12. End if

Fig: 4.1. Nodes in the system

Fig: 4.3. Broadcasting CRQ packets (Voting Phase)

Phase-3 Acceptance Phase:

Fig: 4.2. Identifying filtered nodes (FN) “The node with the highest count will declare itself as
coordinator by passing the CAM (sys-id, voting-count).”
Phase-2: Voting Phase From the second phase whichever the node that retains
“Here the filtered nodes will broadcast CRQ packet to the highest voting count from their CRQ packet (sys-id,
every nodes in the system to attain the votes.” pri-id, voting-count, visitors-id) by taking the
This is our second phase, where the consideration from how many visitors it got it, it will
filtered node (FN) plays an important role in gaining announce itself as the coordinator now.
their votes from the remaining nodes in the system by It will generate the CAM message from its received
visiting them. Through its own CRQ (sys-id, pri-id, input to broadcast by, inputting the voting count so that
voting-count, visitors-id) message filtered nodes (FN) every node in the system will know the node’s majority
are going to broadcast their pri-id and voting-count so count to be the coordinator.
that the interested nodes can input their vote into this
node. Finally, the node with the highest voting-count can
enter into the phase three called as Acceptance phase.
 If the system gets two nodes with the same voting-
count then supported nodes priority will be
summed up to select the node to enter into the
Acceptance phase.
 If two nodes have same pri-id then the voting-count
will be taken into consideration to make that node
to enter into the Acceptance phase.

Algorithm
// To initiate voting:
1. Create message CRQ (sys-id, pri-id, voting-
count, visitors-id)
2. Send CRQ (sys-id, pri-id, voting-count, Fig: 4.4: Broadcasting CAM packets (Acceptance Phase)
visitors-id)
3. // upon receiving a message CRQ
4. If one node has highest voting then

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 901

 Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

D. 3PLE Procedure
1. Initialize the system
V. COMPLEXITY ANALYSIS
2. Phase1: Perform filtration of the nodes:
ER1: Configuration load In this section we have discussed the message and time
ER2: Configuration load complexity of the proposed algorithm.
ER3: Performance
A. Message Complexity
3. Compare each node in the system for the above
combinations and selected will be considered as Best case: In the best case, message will be created from
participants (Filtered Nodes (FN)) for the election. a high priority process which satisfies all the three rules of
4. Phase 2: Broadcast CRQ message packet in the election phase. This leads to message passing involves only
system n number messages to elect itself as a coordinator. Hence,
5. Send CRQ (sys-id, pri-id, voting-count, visitors-id) the message complexity in best case is O(n).
6. { Average case: As in average case, if more than one process
7. If (a node has the highest voting-count) then having the same priorities as like other the number of
8. It is elected as coordinator
required message passing to complete the process of
9. If (two nodes have the same voting-count) then
10. Supported nodes priority will be summed up to election is M, then M<2n. So, the complexity of message in
declare the coordinator average case is also O(n).
11. End if Worst case: The worst-case complexity is also O(n) since
12. Else the processes which initiates the election are filtered in the
13. Two nodes have the same priority then election phase only. Hence the message passing itself gets
14. Voting-count will be taken into consideration. reduced.
15. End else
16. End if B. Time Complexity
17. Phase 3: Broadcast CAM packet in the system The time complexity of the proposed algorithm is O(n).
18. Send CAM (sys-id, voting-count) The process which satisfies the phase-1 of the algorithm can
19. To declare coordinator. send the message to the remaining process in the system.
20. Exit. Each process identifies the coordinator from the CAM
message and hence it takes O(n).
E. Flowchart

Fig.4.5 3PLE Flowchart

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 902

 Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

VI. COMPARISON WITH OTHER ALGORITHMS [5] Amit Biswas and Animesh Dutta, “A Timer Based
In this segment of the paper, we tried to compare Leader Election Algorithm”, IEEE Conferences on
the performance of our proposed methodology for electing Ubiquitous Intelligence & Computing, Advanced and
Trusted Computing, Scalable Computing and
coordinator with respect to the bully and ring algorithm.
Communications, Cloud and Big Data Computing, Internet
[Table-II] show the required no. messages to elect leader for of People, and Smart World Congress
the algorithms both in best case and worst case. (UIC/ATC/ScalCom/CBDCom/IoP/SmartWorld), Toulouse,
France, July 18-21, 2016.
TABLE II. COMPARISON OF PROPOSED WITH BULLY AND RING

Bully Ring 3PLE [6] MohammadReza EffatParvar, Nasser Yazdani , Mehdi

No. (Proposed) EffatParvar , Aresh Dadlani and Ahmad Khonsari,
of Best Worst Best Worst Best Worst
Case Case Case Case Case case(
“Improved Algorithms for Leader Election in Distributed
Proce
sses (n2) (n3) (2n) (2n2) (n) n) Systems” 2nd International Conference on Computer
(n) Engineering and Technology, Chengdu, China, April 16-18,
5 25 125 10 50 5 5 2010.
10 100 1000 20 200 10 10
50 2500 12500 100 20000 50 50 [7] Ehsan Amiri, Hassan Keshavarz andAfshin Shahriari
100 10000 1000000 200 800000 100 100 Fahleyani, “New Algorithm for Leader Election in
200 40000 8000000 400 320000 200 200 Distributed WSN with Software Agents”, IEEE
International Conference on Space Science and
From the above able it is clear that our algorithm is better Communication (IconSpace), Melaka, Malaysia, 1-3 July
than the classical algorithms in terms of message 2013.
complexity.
[8] Behnish Mann and Alex Arvavid, ”Message Complexity
VII. CONCLUSION AND FUTURE SCOPE of Distributed Algorithms Revisited”, International
In this article, we have a proposed 3 phase leader Conference on Parallel, Distributed and Grid Computing”,
election algorithm, on the assumptions that our system has Solan, India, Dec. 11-13 2014.
no failure node, this algorithm helps to filter the process that
[9] Balmukund Mishra, Ninni Singh and Ravideep Singh,
can be nominees for becoming coordinators in the phase-1
“Master-Slave Group Based Model For Co-ordinator
also known as election phase. This helps us to reduce the Selection, An Improvement of Bully Algorithm”,
number of messages in communication. The results show International Conference on Parallel, Distributed and Grid
that the proposed algorithm is better than the existing ones. Computing, Solan, India, Dec. 11-13,2014.
In future we want to exclude the assumption and implement
our algorithm in the real time distributed scenario and share [10] Sheenu Chhabra, Gajendra Tyagi and Ankit
our results. Mundra,”Location Based Coordinator Election Algorithm
inDistributed Environment”, International Conference on
Computer and Computational Sciences (ICCCS), Noida,
REFERENCES India, Jan. 27-29 2015.
[1] H.Garcia Molina, “Elections in a Distributed Computing
System.” IEEE Transactions on Computers, volume: [11] Arghavani, A., Ahmadi, E. ; Haghighat, A.T.,”
C-31, Issue: 1, Jan. 1982, pages: 48 – 59, Sponsored by Improved bully election algorithm in distributed systems,”
IEEE Computer Society. Proceedings of the 5th international Conference on
Information Technology & Multimedia (ICIMU), Kuala
[2] S.Park, Y.Kim and J.S. Hwang, “An Efficient Algorithm Lumpur, Malaysia, Nov. 14-16 2011.
for Leader-Election in Synchronous Distributed Systems,”
proceedings of IEEE. IEEE Region 10 Conference. [12] Sung Hoon Park, Su-Chang Yoo and Jae-Yoeb Lee ,”
TENCON 99. 'Multimedia Technology for Asia-Pacific The Design of an Election Protocol in Wireless
Information Infrastructure', Cheju Island, South Korea, Cellular Networks”, 11th International Conference on
South Korea , Sept. 15-17 1999. Information Technology: New Generations, Las Vegas, NV,
USA, April 7-9 2014.
[3] M.S.Kordafshari, M.Gholipour, M.Jahanshasi, A.T.
Haghighat, “Modified Bully Election Algorithm In [13] Hsu-Chia Cahng and Chi-Chun Lo, “A Consensus-
Distributed System”, Wseas Conferences, Cancun, Mexico, Based Leader Election Algorithm for Wireless Ad Hoc
May 11-14, 2005. Networks”, International Symposium on Computer,
Consumer and Control, Taichung, Taiwan, June 4-6 2012.
[4] Fouad Hanna, Jean-Christopher Lapayre and Lionel
Droz- Bratholet, “Fault Tolerance Management in [14] Orhan Dagdeviren,and Kayhan Erciyes, “A
Distributed Systems : A New Leader- Based Consensus Hierarchical Leader Election Protocol for
Algorithm” , HPCS, Bologna, Italy, July 21-25, 2014. Mobile Ad hoc Networks”, ICCS 2008: 8th International
Conference, Kraków, Poland, June 23-25, 2008.

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 903

 Proceedings of the Third International Conference on Computing Methodologies and Communication (ICCMC 2019)
IEEE Xplore Part Number: CFP19K25-ART; ISBN: 978-1-5386-7808-4

[15] Noman Mohammed, Hadi Otrok, Lingyu Wang, [21] WanFokkink and JunPang, “Simplifying Itai-Rodeh
Mourad Debbabi and Prabir Bhattacharya, “A Mechanism Leader Election for Anonymous Rings”, Electronic Notes in
Design-Based Multi-Leader Election Scheme for Intrusion Theoretical Computer Science, open access, Volume 128,
Detection in MANET”, IEEE Wireless Communications Issue 6, 23 May 2005, Pages 53-68..
and Networking Conference, Las Vegas, NV, USA, 31
March-3 April 2008. [22] S. Vasudevan ,B. DeCleene , N. Immerman , J. Kurose
and D. Towsley, “Leader Election Algorithms for Wireless
[16] Ahcène Bounceur, Madani Bezoui, Reinhardt Euler, Ad Hoc Networks”, Proceedings DARPA Information
Farid Lalem and Massinissa Lounis, “A Revised BROGO Survivability Conference and Exposition, Washington, DC,
Algorithm for Leader Election in Wireless Sensor and IoT USA, USA, April 22-24 2003.
Networks”, IEEE SENSORS, Glasgow, UK, 29 Oct.-1 Nov.
2017. [23] Navneet Malpani, Jennifer L. Welch and Nitin Vaidya,
“Leader Election Algorithms for Mobile Ad Hoe Network”,
[17] SukumarGhosh and ArobindaGupta, “An exercise in Proceeding DIALM '00 Proceedings of the 4th international
fault-containment: Self-stabilizing leader election”, workshop on Discrete algorithms and methods for mobile
Information Processing Letters, Volume 59, Issue 5, 9 computing and communications Pages 96-103, Boston,
September 1996, Pages 281-288. Massachusetts, USA — August 11 - 11, 2000.

[18] S. Vasudevan, J. Kurose and D. Towsley, “Design and [24] Israel Cidon and Osnat Mokryn, “Propagation and
Analysis of a Leader Election Algorithm for Mobile Ad Hoc Leader Election in a Multihop Broadcast Environment”,
Network”, Proceedings of the 12th IEEE International International Symposium on Distributed Computing, DISC
Conference on Network Protocols, 2004. ICNP 2004. 1998: Distributed Computing pp 104-118,
Berlin, Germany, Germany, Oct 8. 2004.
[25] Gang Zhang , Xiaoyan Kuang , Jing Chen and Yu
[19] Awadhesh Kumar Singh and Shantanu Sharma, “Elite Zhang ,” Design and Implementation of a Leader Election
Leader Finding Algorithm for MANETs”, 0th International Algorithm in Hierarchy Mobile Ad hoc Network”, 4th
Symposium on Parallel and Distributed Computing, Cluj International Conference on Computer Science &
Napoca, Romania, July 6-8 2011. Education, Nanning, China, July 25-28 2009.

[20] Tomasz Jurdzinski, Miroslaw Kutylowski, and Jan [26] Ahcene Bounceur, Madani Bezoui, Massinissa Lounis,
Zatopianski, “Efficient Algorithms for Leader Election in Reinhardt Euler and Ciprian Teodorov, “A New Dominating
Radio Networks”, Proceedings of the twenty-first annual Tree Routing Algorithm for Efficient Leader Election in IoT
symposium on Principles of distributed computing Networks”, 15th IEEE Annual Consumer Communications
Pages 51-57, Monterey, California — July 21 - 24, 2002. & Networking Conference (CCNC), Las Vegas, NV, USA,
Jan. 12-15 2018.

978-1-5386-7808-4/19/$31.00 ©2019 IEEE 904