**Jozef Papan 1,\*, Pavel Segec 1, Oleksandra Yeremenko 2, Ivana Bridova <sup>1</sup> and Michal Hodon <sup>3</sup>**


Received: 22 May 2020; Accepted: 15 June 2020; Published: 17 June 2020

**Abstract:** The sprawling nature of Internet of Things (IoT) sensors require the comprehensive management and reliability of the entire network. Modern Internet Protocol (IP) networks demand specific qualitative and quantitative parameters that need to be met. One of these requirements is the minimal packet loss in the network. After a node or link failure within the network, the process of network convergence will begin. This process may take an unpredictable time, mostly depending on the size and the structure of the affected network segment and the routing protocol used within the network. The categories of proposed solutions for these problems are known as Fast ReRoute (FRR) mechanisms. The majority of current Fast ReRoute mechanisms use precomputation of alternative backup paths in advance. This paper presents an Enhanced Multicast Repair (EM-REP) FRR mechanism that uses multicast technology to create an alternate backup path and does not require pre-calculation. This principle creates a unique reactive behavior in the Fast ReRoute area. The enhanced M-REP FRR mechanism can find an alternative path in the event of multiple links or nodes failing at different times and places in the network. This unique behavior can be applied in the IoT sensors area, especially in network architecture that guarantees reliability of data transfer.

**Keywords:** Internet of Things (IoT); ReRoute; Multicast Repair (M-REP)

## **1. Introduction**

The Internet of Things (IoT) model allows the connection and exchange of data between various types of smart devices. These smart devices, usually sensors, can be connected in the Wireless Sensor Network (WSN) [1–4] creating a unique sensor architecture [5,6]. With increasing numbers of sensors in the environment and the importance of measured data, the network platform must guarantee the reliability of connection.

Historically, Internet Protocol (IP) networks have been focused mainly on time-tolerant communication services, such as e-mail, file transfer and access to web content. However gradually, IP networks have evolved into converged platforms supporting several different types of services, including time-consuming and real-time applications such as voice transmission over IP, Internet of Things platform, sensors, streaming and multimedia services [7,8]. These services have higher network performance requirements, such as delay, availability, or packet loss, and are also negatively affected by unexpected link or node failures in the network. In case of network failures, network routing protocols (IGP), such as Open Shortest Path First (OSPF), respond to network failures by flooding topology updates and calculating new routes [7–11]. This process is also known as the network convergence

process. Thus, in a period of network convergence, network routers have outdated information that can cause data loses and outages.

For this reason, IP networks must meet specific qualitative and quantitative parameters in order to ensure an acceptable quality of service, e.g., availability and short repair time after a specific network failure. To do this, network providers must deploy appropriate technologies to ensure uninterrupted customer service. One such significant technology is a group of mechanisms for rapid network recovery, also known as Fast ReRoute (FRR) [12–15]. A key principle of FRR mechanisms is that the backup path for possible failure scenarios is calculated in advance before the failure occurs. This principle presumes that the switching to a precomputed backup path is faster than waiting for the network convergence process to complete. FRR mechanisms that are designed to work on IP networks are known as Internet Protocol Fast ReRoute (IP FRR) mechanisms.

The paper presents an enhanced version of Multicast Repair (M-REP) FRR mechanism (hereinafter EM-REP) that does not require pre-calculation of alternative routes in advance. At the same time, EM-REP is capable of finding an alternative path even in the case of multiple network failures. These features make EM-REP a unique IPFRR mechanism which can be also used in IoT architecture providing reliable connection for various types of sensor networks.

The rest of this paper is organized as follows. Section 2 provides an insight into the issues of Fast ReRoute, the purpose, and terminology used. Section 3 provides state-of-the-art analysis, where existing FRR solutions and identified problems are discussed. Section 4 presents the specification of the original M-REP FRR mechanism and Section 5 proposes its enhanced version, the EM-REP FRR mechanism. Section 6 focuses on the evaluation of the EM-REP mechanism and compares its features with the predecessor, the M-REP algorithm, as well as with the other FRR solutions. Finally, Section 7 presents the conclusions and directions for future research work.

#### **2. The Fast ReRoute**

One of the main reasons why Fast ReRoute mechanisms have started to be used is that the process of network convergence after a network failure usually takes a longer time than that expected by the network provider.

In the context of routing, the convergence state of a network is the state in which each network router has up-to-date, complete and consistent routing information [16,17]. This means that each router is informed of the current network status (up-to-date), of each available network (complete), and each router has chosen the optimal successor/next-hop for each network according to a common criterion (consistent). The convergence process is the operation of each individual routing protocol process to achieve a state of convergence. The time required to complete the process of network convergence depends on various factors, such as the number of devices, especially network routers; topology complexity; the type of routing protocol used (link-state or distance vector); and its individual operational parameter values (like update or hello timers). The time required to finalize the network convergence process usually ranges from hundreds of milliseconds up to a few seconds. The network convergence process may consist of several subprocesses, such as failure detection, response, the distribution of updated routing information, recalculation and, finally, the installation of new routes [7,18–20]. The resulting time of network convergence is the sum of durations of each individual subprocesses.

First, from the router point of view, there is a need for mechanisms that perform failure detection. Failure detection is the time that the router's operating system (OS) needs to detect a failure, i.e., a failed link (network interface) or unavailable neighbor router. Interface failure detection may take up to several milliseconds on the physical layer of a router. The detection of neighbor router unavailability depends on the type of routing protocol used. This failure detection may take from tens of milliseconds when a link state protocols are used (OSPF/IS-IS hello mechanisms), or it may take up to tens of seconds in the case of distance vector routing protocols (RIPv2). During this period, the packets affected by the failure are permanently lost due to outdated routing information and resulting incorrect routing

decisions. Next, the local router response to a link failure; it generates and distributes topology/routing updates that reflect actual state. The duration depends on actual load conditions of each individual router. The distribution of information to other routers is required to inform other routers about the situation and starts within 10 ms to 100 ms for each affected next-hop router [18–20]. All routers that have received actual routing information must recalculate their routing tables. The recalculation of routing tables depends mainly on the size of the network and the amount of topological information. This may take a few milliseconds for link-state routing protocols that use the Dijkstra algorithm. After the recalculation is complete, routers install new routes and update their routing tables. Again, this mainly depends on the type of router and the number of prefixes that were affected by the network failure.
