Analytical decisional model for latency aware publish/subscribe systems on MANET
Introduction
Mobile ad hoc networks (MANET) (Basagni, Conti, Giordano, Stojmenovic, 2004, Zhang, Zhou, Yu, 2006) are a set of nodes, communicating meaning wireless channel without any centralized control. Each node may be either a data sender, a data receiver or an intermediate router. The main characteristic of MANET is the frequent mobility of its nodes leading to some dynamic changes of topology. Nodes mobility in such networks introduces possible disconnections between adjacent nodes, and more generally quality of service (QoS) degradation issues (such as long delay, low throughput) that do not fit well with QoS requirements of QoS sensitive applications such as multimedia (audio, video) applications.
At the routing (i.e. third layer in the OSI Network) layer, different QoS oriented routing protocols have been introduced aiming at finding the corresponding path between the source and the destination. Routing protocols (Mammeri, 2010, Yanjing Sun, Sun, Hu, 2014, Sridhara, Baskaranb, Chandrasekara, 2013, Bara, Mandala, Singhb, 2013, Chena, Wenga, Kuob, 2010) have to be able to deal with crash and breakage affecting link and re-route messages, whenever any faults affecting the intermediate nodes on the route due (for instance) to mobility is susceptible to make inadequate the end-to-end route. However, if there is no such a path between end-nodes, the targeted QoS cannot be provided. More generally, maintaining a QoS level that matches the application level QoS requirements is a very complex task.
Hence, it becomes very necessary for various applications to have a strength QoS mechanism support especially those requiring a special need for QoS like crisis management applications. Thus, above the routing layer, these applications introduced the middleware layer in order to solve problems unmanageable by the previous one.
At the middleware layer, some communication paradigms have been evaluated in order to find a way to connect mobile entities in the mobile network such us publish/subscribe, remote procedure calls, and tuple spaces paradigms (Collins, Bagrodia, 2012, Denkoa, Shakshukib, Malikc, 2009).
Publish/subscribe has been well evaluated in a qualitative comparison between these communication patterns.
In this paper, we aim to provide a QoS management at the middleware layer while assuming a best effort flux at the routing layer. We also use of the publish/subscribe paradigm to communicate mobile entities in MANET.
A publish/subscribe (pub/sub) (Carzaniga et al., 2001) communicating system is a decoupled interaction between publishers (or producers) and receivers (or consumers). Publishers provide messages to be delivered to consumers whose previously subscribe for that event. This is ensured by an intermediate component named broker (or dispatcher or channel manager). The broker is the most important component of a pub/sub communication infrastructure. Two of its main functions are: (1) to perform the matching between the incoming events and the subscriptions previously registered, and (2) to send incoming events to the receivers that have expressed an interest in the events. Let us note that a pub/sub communication system may include several brokers between final senders and receivers.
Applied at the middleware level (i.e. between the Application and the Transport layers of the OSI reference model), the pub/sub paradigm is promising for MANET thanks to its decoupling nature.
Several research efforts have been already accomplished in the study of pub/sub paradigm to enhance QoS in MANET and to tolerate failures (Cugola, Picco, 2006, Denkoa, Shakshukib, Malikc, 2009, Kazemzadeh, Jacobsen, 2009, Malekpour, Carzaniga, Pedone, 2011, Esposito, Cotroneo, Russo, 2012, Esposito, Cotroneo, Russo, 2013, Maye, Brunie, Coquil, Kosch, 2012, Zhao, Wu, 2013, Esposito, Platania, Beraldi, 2014, Hoffert, Mack, Schmidt, 2009). In this direction, adopting autonomic and self-reconfiguring system have gained raising interest during last years. Indeed, the challenge is that the system have to be ready to recover failures beyond human intervention. Autonomic systems are structured following the IBM autonomic process (Jacob et al., 0000) into four steps namely monitoring, analysis, plan and execution. Significant efforts have been dedicated in each of these phase.
Indeed, in the execution step, faults recovery schemes have been proposed in order to restore system functioning (Kim, Karenos, Ye, Reason, Lei, Shagin, 2010, Chen, Schwan, 2005, Avvenuti, Vecchio, Turi, 2005, Cugola, Picco, 2006, Denko, Shakshuki, Malik, Pankajakshan, Rao, 2012). The starting point of these approaches relies basically on the hypothesis that QoS maintenance might be improved thanks to a dynamic location of the brokers on the nodes. Our starting point is similar and considers that the broker location may be driven by two main criteria: the mobility of the hosting nodes, and the QoS (here the latency) observed between nodes hosting adjacent brokers along the data path. When mobility increases, we assume that there is a serious risk of QoS degradation between the broker and its neighbor brokers, and that the broker location has to be reconsidered. Similarly, when the observed latency between two adjacent brokers increases, we assume that this involves a problem (e.g. node congestion) that is time to reconsider the broker location.
As a solution to the mentioned faults problem, reconfiguration actions be elaborated in order to safeguard system survivability according to each QoS degradation cause. In fact, adaptation can be achieved through moving a nonstable broker to another location (i.e node) in the field providing a better QoS level whenever failure is due to a node mobility. It is also possible to split an overloaded broker into equilibrate brokers load (Abid et al., 2013).
In our study, the general approach is supposed to be performed following the autonomic paradigm. Our contribution lies in the monitoring and analysis steps of the MAPE loop. Monitoring consists in computing latency values taken by events between neighboring brokers. The analysis consists first in predicting/detecting QoS degradation and then in looking for the cause of the degradation which we would call the diagnosis.
The previously cited contributions that target the same issues are based on the comparison of QoS parameters values with fixed thresholds and are restricted to a specific subscription language and a specific topology. Otherwise, the approaches proposed in the literature impose a specific form of analysis (i.e. either reactive of predictive), which cannot be suitable for all types of applications. Our approach is different in the way we propose to base the analysis on the dynamic thresholds enabling us to take into account network dynamicity. The analysis step is driven thanks to an analytical model combining reactive, predictive and hybrid analysis which allows meeting the different kinds of application requirements.
The new model provides then three types of analysis. The developer must then choose the form of analysis considering his application: (1) a reactive analysis (Lahyani et al., 2012c) which is best adapted to multimedia applications where users need to have a reliable video transmissions. (2) A proactive analysis (Lahyani, Makki, Chassot, 2012, Lahyani, Gassara, Jmaiel, Chassot, 2012) which fits the best to critical applications such as crisis management. A failure in such applications must be predicted since it may lead to a catastrophe. (3) Hybrid analysis (Lahyani et al., 2012) which is a mixture of the reactive and the predictive forms of analysis. The new analytical module is more generated since it is not specific to a particular application. It may be applied to different types of problems (multimedia, crisis management applications…).
Our customizable approach relies on mathematical models. Indeed, the reactive analysis which is more suitable for video streaming multimedia applications, approximates the monitored latency values with the Gumbel (Giorgi and Narduzzi, 2008) distribution towards computing thresholds. Latency is then compared with adjustable thresholds updated using the Exponential Weighted Moving Average Formula.
Proactive analysis is better appropriate to crisis management applications where failures should be omitted since it leads to catastrophe. The predictive approach forecasts failures by adopting the Auto Regression (Knight, 1989) as well as the Auto Regressive Integrated Moving Average Method (Ruey and Tsay, 2008). However, the hybrid analysis is a combination between the two pre-described forms of analysis. Once QoS degradation is noticed, diagnosis step involved in the analysis module is carried out using the correlation method1 in order to identify failure causes. This is very important since adaptation is carried out considering failures causes. Indeed, the proposed decisional module creates opportunities to repair the system.
In order to evaluate the proposed analysis models, simulation was achieved on an ad hoc network using the Jist/Swans Simulator (Barr, 2005). In this paper, new experimental tests were achieved considering different mobility conditions regarding the pause time parameter related to the mobility model. Thus, we realize experiments in medium, low and high mobility scenarios. The Results prove that the proposed analytical framework can efficiently detect/predict failures and save system breakdowns.
Besides, this paper show the benefits gained by introducing the proposed approach in the considered publish/subscribe system when compared with other scenarios integrating analysis module relying on fixed thresholds.
The remainder of this paper was organized as follows. The second section introduced the related work to the topic. A detailed description of the proposed decisional model for latency aware pub/sub systems on MANET was detailed in Section 3. Thereafter, in Sections 4 and 5, some evaluations using simulation experiments were described. Finally, we drew our conclusions before suggesting some perspectives in Section 6.
Section snippets
State of the art
Significant research study has been devoted to the provision of QoS guarantees in MANETs at the routing layer (the third layer in the OSI model). Considering some QoS parameters, efficient routing protocols in MANET aim to find available paths for a connection based on QoS requirements applications (Sridhara, Baskaranb, Chandrasekara, 2013, Bara, Mandala, Singhb, 2013, Chena, Wenga, Kuob, 2010). To support such QoS requirements, some QoS parameters have been incorporated in route finding and
The proposed analytical model
In this paper, we described our analytical model involved in each broker at the middleware layer, aiming at supervising and analyzing the system facing MANET dynamic behavior. The primary target of the proposed approach was to overcome link faults and raise the opportunities of meeting QoS applications requirements.
The proposed model is part of a general approach that supposed to be performed following the autonomic paradigm defined by IBM (Jacob et al., 0000), that maps all of the monitoring,
Simulations
The JiST/SWANS simulator (Java in Simulation Time/Scalable Wireless Ad hoc Network Simulator) (Barr, 2005) was used to simulate a dynamic network environment. SIENA (Carzaniga et al., 2001) (Scalable Internet Event Notification Architecture) was also resorted to as a pub/sub-content based system to which we added our monitoring and analysis modules. During simulation, the latency between two neighboring brokers called B1 and B2 was measured at each delivery event.
Simulation involves three
System performance
For a better assessment of the involved tradeoffs, let us consider the system availability (Torell and Avelar, 2010) as a parameter to measure the system performance. In our context, availability is defined as the probability that the system will perform its intended function during a specified period of time under the stated conditions. So the system availability is equivalent to link availability by adopting reliable nodes. Thus, mathematically, availability is estimated using the following
Conclusion
This paper proposes a latency aware pub/sub system on MANET involving monitoring and analysis modules. Our goal was to provide the best adapted QoS in the system according to the available resources and thus to overcome link failures. A customizable model proposing reactive, predictive as well as hybrid analyses is the proposed to give strong support to QoS requirements and eliminate system breakdown situations. As a future perspective, a test bed will be set up to embed the existing analytical
Imene Lahyani obtained her diploma of engineer in Computer Science from the National School of Engineering of Sfax-Tunisia in 2007 and her M.S. degree (DEA) from the same school in 2008. Currently, she is a PhD student at the National School of Engineering of Sfax-Tunisia. Her current research areas include software engineering of distributed systems, self-adaptive systems, autonomic middleware. She published serveral papers in high ranked international conferences. She is assistant in the
References (50)
- et al.
Survey on reliability in publish/subscribe services
J. Comput. Netw.: Int. J. Comput. Telecommun. Netw. Arch.
(2013) - et al.
A cross-layer middleware for context-aware cooperative application on mobile ad hoc peer-to-peer network
J. Syst. Softw.
(2014) Tutorial on maximum likelihood estimation
Sci. Dir.
(2003)- et al.
Exploration de l’espace des configurations architecturales guidé par la qds des systémes publier/souscrire sur manet
Proceedings of the 7éme édition de la Conférence francophone sur les Architectures Logicielles (CAL)
(2013) - et al.
A cross-layer approach for publish/subscribe in mobile ad hoc networks
Mobile Agents for Telecommunication Applications
(2005) - et al.
QOS of manet through trust based aodv routing protocol by exclusion of black hole attack
Proc. Technol.
(2013) - Barr, R.H.Z.R.R., 2005, Scalable wireless ad hoc network simulation, handbook on theoretical and algorithmic aspects of...
- et al.
Mobile Ad Hoc Networking
(2004) - et al.
A QoS-Oriented Reconfigurable Middleware for Self-Healing Web Services
Proceedings of the International Conference on Web Services (ICWS), September 23–26, 2008
(2008) - et al.
Design and evaluation of a wide-area event notification service
ACM Trans. Comput. Syst.
(2001)
Opportunistic overlays: efficient content delivery in mobile ad hoc networks
Proceedings of the ACM/IFIP/USENIX 2005 International Conference on Middleware
Signal strength based routing for power saving in mobile ad hoc networks
J. Syst. Softw.
A quantitative comparison of communication paradigms for manets
Proceedings of the 6th ACM International Conference on Distributed Event-Based Systems, Lecture Notes of the Institute for Computer Sciences, Social Informatics and Telecommunications Engineering, vol. 73
Enhanced cross-layer based middleware for mobile ad hoc networks
J. Netw. Comput. Appl.
A tutorial on reliability in publish/subscribe services
Proceedings of the 6th ACM International Conference on Distributed Event-Based Systems
Reliable and timely event notification for publish/subscribe services over the internet
IEEE/ACM Trans. Netw.
A survey of online failure prediction methods
ACM Comput. Surv
Reds: a reconfigurable dispatching system
Proceedings of the 6th International Workshop on Software Engineering and Middleware
Detection of anomalous behaviors in networks from traffic measurements
IEEE Trans. Instrum. Meas.,
Using machine learning to maintain pub/sub system QOS in dynamic environments
Proceedings of the 8th International Workshop on Adaptive and Reflective MIddleware
A fast and robust content-based publish/subscribe architecture
Proceedings of the Seventh IEEE International Symposium on Network Computing and Applications, 2008. NCA ’08
Fail-aware publish/subscribe
Proceedings of the Sixth IEEE International Symposium on Network Computing and Applications, 2007. NCA 2007.
Cited by (1)
Mobility prediction for random walk mobility model using ARIMA in mobile ad hoc networks
2022, Journal of Supercomputing
Imene Lahyani obtained her diploma of engineer in Computer Science from the National School of Engineering of Sfax-Tunisia in 2007 and her M.S. degree (DEA) from the same school in 2008. Currently, she is a PhD student at the National School of Engineering of Sfax-Tunisia. Her current research areas include software engineering of distributed systems, self-adaptive systems, autonomic middleware. She published serveral papers in high ranked international conferences. She is assistant in the National School of Engineering of Sfax-Tunisia since October 2012. More details are available on her home page: http://www.redcad.org/members/imene.lahyani/.
Mohamed Jmaiel obtained his diploma of engineer in Computer Science from Kiel (Germany) University in 1992 and his Ph.D. from the Technical University of Berlin in 1996. He joined the National School of Engineers of Sfax (Tunisia) as Assistant Professor of Computer Science in 1995. He became an Associate Professor in 1997 and full Professor in January 2009. He participated to the initiation of many graduate courses at the University of Sfax. His current research areas include software engineering of distributed systems, formal methods in model-driven architecture, component oriented development, self-adaptive and pervasive systems, autonomic middleware. He conducted many research projects and published more than 120 regular and invited papers in international conferences and journals, and has co-edited four conferences proceedings and three journals special issues on these subjects. He organized and co-chaired the program committees of many international conferences, like MCSEAI’2004, CRiSIS’2009, NOTERE’2010, OPODIS’2010, and WETICE’2013. He was the Director of the National Engineering School of Sfax (ENIS), from 2011 to 2014. Currently, he is the Director of the research center on TIC at the Technopark of Sfax. More details are available on his home page: http://www.redcad.org/members/jmaiel/.
Christophe Chassot received the Engineering degree and the M.S. degree (DEA) in computer Science from the National Polytechnic Institute of Toulouse (INPT) in 1992, and the Ph.D. degree in Computer Science from INPT, in 1995. Since 1996, he has a lecturer position (Maitre de Conférences) at the National Institute of Applied Sciences of Toulouse (INSA). He is also associate Researcher at the Laboratory of Analysis and Architectures of Systems (LAAS), a laboratory of the National Center of the Scientific Research (CNRS). His main fields of interest include QoS-oriented self adaptive Transport architectures, and signaling for multi-domains DiffServ Internet.