The Intelligent Router-Assisted (IRouter) Power Saving Medium
4.2 The IRouter Algorithm
In the IRouter algorithm, the MNs within the network topology can be classified based on their functionalities, i.e. N = {S, D, R, I} = {Si, Di, Ri, Ii| 1 ≤ i ≤ n}. S represents the set of source MNs that initiate data packet transmission; D contains the destination MNs that only receive data packets; R encloses the MNs that redirect the data packets between nodes Si and Di; I is the set of inactive MNs. It is noted that the router node Ri for one route is also capable of being the data source or destination for other routes within the network. The functionality of each MN will be defined based on its Partial Route Table, P RT , which will be explained in the following.
The concept of the proposed IRouter scheme is based on the high probability that same routes will be utilized for delivering more than a single data packet. Partial routing infor-mation, including the previous and the next hopping nodes, is recorded in a cache memory within the router MNs, i.e. Ri.
Fig. 4.2 shows an example of network topology that utilizes the proposed IRouter scheme.
There are three different routes (Γ1, Γ2, Γ3) that are partially illustrated in the network topology in consideration, i.e. Γ1 = {. . . , Rh, Ri, Rj, Rk, Rl, . . .}, Γ2 = {Si, Rj, Rx, . . .}, Γ3
= {. . . , Ry, Di}. In order to justify the proposed scheme, the route between nodes Ri, Rj, and Rkand their associated transmission range circles are the primary concerns in this thesis.
It is noted that the determination of the multi-hop sequences within a route are conducted by the routing layer protocols. The MAC layer algorithm within a router node is aware of
Ri
Figure 4.2: The Network Topology with the Proposed IRouter Scheme the information of its next delivering MN only if the following steps are fulfilled:
1. the data packet has arrived in the MN;
2. based on the information obtained from the received data packet, the routing protocol determines its next forwarding MN within the route;
3. the information is forwarded to the MAC layer algorithm to schedule for the data trans-mission.
The design concept of the proposed IRouter scheme is to predict the next hopping MN before the actual arrival of the data packet. During the time interval for ATIM hand-shaking, the predicted next delivering MN (obtained from the information recorded in the P RT ) is requested to be in the awake state for potential data transmission. The Par-tial Route Table, P RT = hEN T1, EN T2, . . .i, created within the cache memory of each MN is utilized to record partial routing information. The three fields within each entry EN Ti = (U pM N, DownM N, T T L) of the P RT is explained as follows:
1. U pM N : The first field within each P RT entry indicates the upstream MN within the route.
2. DownM N : The second field in each P RT entry designates the downstream MN of the route.
3. T T L: The Time-To-Live (T T L) field is utilized to represent the freshness of the corre-sponding P RT entry. The smaller T T L value indicates the more up-to-date information is recorded in the previous two fields. While T T L is equal to zero, the P RT entry con-tains the newly updated upstream and downstream information.
The P RT of Rj is obtained as P RTRj = h(Ri, Rk, 0), (Si, Rx, 1)i, which indicates that Rj is a router node for the two different routes, Γ1 and Γ2 , as the example illustrated in Fig. 4.2. The first entry illustrates that node Rj has Ri as its upstream MN; while Rk is the corresponding downstream MN. It is noted that T T L = 0 represents that the most up-to-date information is recorded in this entry. The second entry shows that node Rj also has Si as its upstream MN in the route Γ2; while Rx is the downstream MN (with T T L = 1). The P RT within the source node Si is obtained as P RTSi = h(φ, Rj, 1)i, where φ indicates that the source node Si does not have a upstream MN within the route Γ2 . The P RT s within other MNs (as shown in Fig. 4.2) can also be observed in the similar manners.
The proposed IRouter scheme consists of two processes, including the P RT construction and the route prediction processes. The route prediction process forecasts the next hopping MN based on the information obtained from the P RT construction process. The proposed algorithm is described in the following two subsections:
4.2.1 P RT Construction Process
As shown in Figs. 4.1 and 4.2, the data packets are assumed to be transmitted from node Ri, via Rj, to Rk. All the MNs within the neighborhood of node Ri are synchronized at the beginning of the BI and stay in the awake state during the entire ATIM window. Other MNs within the communication range (i.e. nodes Rh, Rj, and Si) may intend to compete the time slot for transmitting its packets in queue. It is assumed that node Ri wins the channel contention and sends out an ATIM frame requesting for the data transmission. In this example, the destination node ID within this ATIM frame is designated as node Rj. Node
Rj replies with an ATIM-ACK frame to Ri; while other MNs discard this ATIM frame after receiving it (e.g. nodes Rh, Rj, and Si). Node Rj records the ID of the upstream node in one of its P RTRj entries (i.e. the ID of node Ri as shown in Fig. 4.2). The T T L value in the corresponding entry is reset to zero. After receiving the ATIM-ACK frame from node Rj, node Ri also records the ID of Rj as the DownM N in one of its P RTRi entries.
After the ATIM window is finished, node Ri initiates the transmission of data packet to node Rj. In additions, only nodes Ri and Rj are in the awake state after the ATIM window as shown in Fig. 4.1. For power-saving purpose, the power mode of other MNs are in the sleep state, which will not be able to either transmit or receive data packets until the next BI starts. Since the data packets are assumed to be routed from node Ri, via Rj, to Rk, the transmission from node Rj to Rk follows similar procedures as that from node Ri to Rj. As can be seen from Fig. 4.2, several fields within the P RTRj and P RTRk are filled (i.e. the ID of Rk is recorded in the DownM N field in the corresponding P RTRj entry; the ID of Rj is inserted into the U pM N field in the corresponding P RTRk entry). The power states within the P RT construction process for transmitting data packets can also be observed from the timing diagram as in Fig. 4.1.
4.2.2 Route Prediction Process
The effectiveness of the proposed IRouter scheme can be examined from the route prediction process of the algorithm. The concept of the algorithm is based on the high probability that the same route should be utilized for delivering more than a single data packet. For instance, there is high possibility to deliver the remaining data packets using the same route from node Ri, via Rj, to Rk. After the P RT construction process, some of the MNs pertain certain information in their P RT s as shown in Fig. 4.2. Within the routing layer protocol, it is assumed that another data transmission from node Ri, via Rj, to Rk is decided. The IRouter algorithm of node Ri is requested for the packet delivery to node Rj at the beginning of the third BI (as shown in Fig. 4.1). Node Risends out an ATIM frame, which is destined to node Rj, to its neighborhood MNs. After receiving the ATIM frame from node Ri, Rj replies with
an ATIM-ACK frame to Ri for acknowledgement.
One of the major characteristics of the proposed IRouter scheme is that each MN will verify its P RT after sending out the ATIM-ACK frame to the corresponding requestor. As shown in the route prediction process as in Fig. 4.1, node Rj continues to verify if there is any routing information existed within its P RTRj after sending the ATIM-ACK frame to node Ri. Based on the information from the P RTRj(Ri, Rk, 0) entry within node Rj, it is recognized that Rk has high possibility to be the next hopping node within the route. Node Rj will initiate an ATIM frame to node Rk within the same ATIM window as shown in Fig.
4.1. It is noticed that there can be many MNs competing for sending their ATIM frames (e.g.
nodes Rh, Rj, and Si as in Fig. 4.2) within the same ATIM window.
After receiving the ATIM frame from node Rj, Rk replies with an ATIM-ACK for hand-shaking. Similar route checking mechanism continues at node Rk to determine if there is another next hopping node within the route. Based on the information obtained from the P RT of each MN, additional pairs for ATIM/ATIM-ACK handshaking can be delivered within the same ATIM window as the length of the window permits.
After the time for the ATIM window elapse, nodes Ri, Rj, and Rk are requested to be in the awake state; while other MNs are set to the sleep state. Within the remaining time interval of the same BI, nodes Riand Rj conduct Data/Data-ACK handshaking for data transmission.
After receiving the data packet from node Ri, the actual packet transmission from node Rj to Rkwill be verified by the routing layer protocol. The IRouter scheme of node Rj forwards the data packet to its upper layer routing algorithm for the determination of the next hopping MN. Assuming that PRdownj hP, N i represents the common packet header acquired from node Rj’s routing algorithm down to the IRouter MAC scheme, where P indicates the previous hopping node; and N stands for the next hopping node as shown in Fig. 4.3. If it is observed that P = Ri, there can be two different cases occurred depending on the values within the P RTRj(Ri, Rk, 0) and PRdownj hP, N i:
• If N = Rk, it indicates that the routing decision within node Rj determines to forward the data packet from Rj to Rk. Node Rj can perform the packet delivery to node Rk
Figure 4.3: The IRouter MAC Scheme Will Check the P RT and PRdownj hP, N i
within the same BI since Rk is the predicted DownM N with awaked power state. The same prioritized contention mechanism as in (??) is utilized to ensure that the data packet is delivered with relatively higher priority.
• If N 6= Rk, the routing decision within node Rj designates that the data packet from node Ri should be forwarded to node N , which is different from the DownM N field of the P RTRj(Ri, Rk, 0). There will be no further data transmission from node Rj to either node Rk or N within this BI since the power state of node N is unknown at this point. The P RTRj(Ri, Rk, 0) will be modified as P RTRj(Ri, N , 0) to reflect the new update from the routing decision. Until the next BI starts, the ATIM/ATIM-ACK handshaking will be performed between nodes Rj and N to wake up both MNs for data transmission.
It is noticed that there are chances that the IRouter algorithm within node Rj may not receive indication from its routing layer protocol for a pre-specified time interval. It implies that the data packet should not be forwarded to node Rkbased on the routing decision. The P RTRj(Ri, Rk, 0) will be changed to P RTRj(Ri, ∅, 0), which indicates that no further data forwarding is required.
As was mentioned previously, the T T L value in each P RT entry is utilized to show the
freshness of its associated partial route. The update scheme for the T T L value is as follows.
If it is found that the next hopping MN matches with the predicted DownM N during the route prediction process (i.e. N = Rk), the T T L value will be reset to zero to represent the up-to-date information in the corresponding P RT entry. The T T L value in each entry will be increased by one at the beginning of each BI. If the T T L value is greater than a pre-specified value η, the corresponding P RT entry will be discarded. It indicates that the information stored in the P RT entry is comparably out-of-date for the MN to adopt. Based on the route prediction mechanism of the proposed IRouter scheme, the data packet can be routed between more than two MNs within a single BI. In the next section, the effectiveness of the IRouter algorithm will be examined in simulations.