*2.1. The Principle of Fast ReRoute*

For the proper understanding of the FRR technology, there is a need for terminology that denotes routers with special meaning for FRR mechanisms [21–25]. Here, we introduce the terms using the following simplified network topology described in Figure 1. The source router (S) is a router that has detected a link or neighboring router failure and then activates a locally implemented FRR repair mechanism. In other words, router S is actively involved in FRR repair (Figure 1). This router is also called Point of Local Repair (PLR). The destination router (D) is the destination router of the original data flow. Routers N1, N2, N3 and others are specific routers that are used as an alternative next router (hereafter referred to as next-hop router) for a specific FRR alternative path. R (R1) is a router that is not actively involved in FRR repair.

**Figure 1.** The principle of Fast ReRoute (FRR).

However, before starting the main FRR process, an administrator must set up protected links or prefixes that are managed by the router. Subsequently, the FRR mechanism pre-calculates an alternate next-hop router to be used in the event of a protected link or prefix failure. This is called Phase Zero (preparation). FRR can then proceed further through the following phases [26–28]:


• Phase Four: The routing protocol completes the necessary routing information update. As the next step, the FRR mechanism is deactivated and the routing process is taken over by the routing protocol.

Once the update of routing information is completed, the deactivation of the FRR mechanism can be accomplished in several ways. One method used is to apply a hold-down timer. This timer should be set to a minimum time necessary to complete the network convergence process. After this timer expires, the temporary routing information installed by the FRR mechanism is removed and the FRR mechanism is subsequently deactivated [26].

The main advantage of the Fast ReRoute mechanism is that it offers several times faster network transmission recovery than a traditional routing protocol may achieve (OSPF). The average repair time of actual FRR mechanisms is up to 50 ms [26,29–31].

#### *2.2. Precomputation Approach of Fast ReRoute*

A key feature common to Fast ReRoute mechanisms is that they calculate the backup path in advance and therefore offer faster network recovery [32,33]. The precalculated backup path in the FRR terminology is also referred to as a precomputed backup path [34,35]. To ensure correct network recovery, the backup path cannot pass through the failure point. Depending on the FRR mechanism, a given router may also calculate several backup paths. When calculating and installing a pre-calculated alternative route, each router decides independent of other routers. The principle of precalculated alternative routes is currently used by all FRR mechanisms. This proactive approach is an important factor in minimizing the time required for fast network recovery after failures [32].

#### **3. Related Works**

In the IoT area, several existing solutions dealing with rerouting have been proposed. In paper [36], a new approach of jamming attack tolerant routing using multiple paths based on zones is presented. The proposed scheme in that paper separates the system network into specific number of zones and directs the candidate forward nodes of neighbor zones. After detecting a specific attack, detour nodes in the network determine zones for rerouting and detour packets destined for victim nodes through forward nodes in the decided zones.

In work [37], the authors present a detailed review of IoT sensing applications in WSN and the difficulties and challenges that need to be overcome. Some of these challenges are fault tolerance, the effectiveness of the energy harvesting, communication interference, cost feasibility, and an appropriate integration of these elements.

At present, there are many unicast IPFRR mechanisms that differ in the way that alternative routes are calculated. Three of the most common and widely used IPFRR mechanisms are the Equal-Cost Multi-Path (ECMP) [30,31], Loop Free Alternates (LFA) [24,31,38] and its extended version, Remote LFA (RLFA) [30,39].

The LFA mechanism calculates alternative routes based on conditions that consider metrics for each next-hop router. These conditions ensure that if a packet is redirected to this alternate next-hop router (that has met the conditions), the router delivers the packet to the destination over a longer path that is still loop-free and bypasses the network failure. The Remote LFA is an improved version of the original LFA. The idea of Remote LFA is to use a tunneling mechanism from the source router S to the remote LFA router. The tunnel is used to bypass the part of the network that, in the event of an error, would route packets (not tunneled) from the affected site back to the source router S or would forward them through a failed link or router. The RLFA router may be a few hops away from the source router S.

Other mechanisms, although less common, are Multiple Routing Configurations (MRC) [11,30,40–44] and Not-Via Addresses [35,45]. Furthermore, there are tunneling-based mechanisms, such as Maximally Redundant Trees (MRT) [46,47], and, finally, IPFRR mechanisms based on alternative trees [22,48–50]. IPFRR mechanisms such as Not-Via Addresses, Multiple Routing Configurations and Maximally Redundant Trees can provide protection that is close to 100% of repair coverage [26,51]. The main challenge of these IPFRR mechanisms is the complexity of internal algorithms that calculate alternate paths.

There is also another FRR group of mechanisms that focus on the protection of multicast communication [52,53]. In general, these solutions utilize precomputed multicast disjoint trees. Examples of these mechanisms are Multicast Only Fast Re-Route (MoFRR) [54] and Bit Index Explicit Replication-Traffic Engineering (BIER-TE) [51,55].

We have been analyzing and researching FRR mechanisms for several years [21,56–61]. Based on the obtained results, we can summarize the most significant properties of the existing FRR mechanisms in Table 1.


**Table 1.** FRR mechanisms: features comparison.

Notes: ECMP—Equal-Cost Multi-Path Routing; BIER-TE—Bit Index Explicit Replication—Traffic Engineering; LFA—Loop-Free Alternate; MoFRR—Multicast-Only Fast Reroute; MRC—Multiple Routing Configurations; MRT—Maximally Redundant Trees; TI-LFA—Topology Independent Loop-Free Alternate.

#### *3.1. Problem Formulation*

In analyzing the FRR mechanisms mentioned above, several issues have been identified. We can classify them into the three basic problem areas, which are briefly introduced in the following subsections.

#### 3.1.1. Cost-Based Calculation of Alternative Route

The majority of existing FRR mechanisms, such as LFA [24,31,62], Remote LFA [14], Directed LFA [63], ECMP [23], MRC [11], and MRT [46,47], calculate an alternative backup route according to link metrics. Alternative routes are usually calculated using a Dijkstra SPF algorithm, which calculates the route path as the minimal total cost of each individual link. The main problem with this type of calculation is that a valid alternative route can only be calculated if the internal algorithm of the FRR mechanism is able to find the correct alternative route according to specific metric conditions. In other words, there are topologies or situations where one FRR mechanism can find an alternative path but another FRR mechanism is unable to do so. If link costs exceed mathematical conditions, the FRR mechanism cannot find an alternative route, even if the alternative route physically exists. The positive effect of the cost-based FRR mechanisms is that they guarantee the calculation of the most advantageous alternative route in the event of a failure. On the other hand, it should be noted that they depend on the correct cost of links in topology. Therefore, there is a need for an FRR algorithm that is able to find an alternative route without cost-based calculation of an alternative route.

#### 3.1.2. Single Failure Recovery

Mechanisms such as Remote LFA [39], Directed LFA [63], and Not-Via Address [45] are designed to be able to protect networks only in the event of a single failure. In situations where more than one failure occurs, these FRR mechanisms cannot create an alternative path and to reroute affected traffic around the failed element in the network. Therefore, packets could be lost in this situation, as the mechanism was not designed to account for more than one point of failure. This is sometimes identified as a limitation of these mentioned FRR algorithms.
