The symbol
ri represents the time at which all the flows
Fi are released for scheduling. The flow
Fi that arrives at the source cluster head is denoted by
. The notation
denotes the flow
Fi arriving at the destination cluster-head. The notation for the node that arrives at the destination node is represented by
. We divide the time slots to avoid interference. Therefore, the proposed scheme is represented by assuming IntraSend, InterComm, IntraRecv as 1, 2, 1 respectively shown in
Figure 1. The minor frame in
Figure 1 shows the number of time slots per frame. IntraSend means the scheduling algorithm which will schedule the flows from the source node to the source cluster head. InterComm represent the scheduling algorithm that will schedule the flows from the source cluster heads to the destination cluster heads. IntraRecv represents the scheduling that will schedule the flows from the destination cluster head to the destination node.
H1,
H2,
H3,
H4,
H5 represent cluster head
1,
2,
3,
4 and
5 respectively as shown in
Figure 2. Transmission
uv means that node
u can send the data to node
v.
Procedure of the Proposed Scheme
The proposed scheme schedules the flows from the source to destination node by using three scheduling algorithms, namely IntraSend, InterComm and IntraRecv. In the first scheduling algorithm, the flows are scheduled from a source node to the source cluster head. Cluster heads are scheduled from source to destination in the second scheduling algorithm. Finally, in the third scheduling algorithms, the flows are scheduled from destination cluster head to destination node. As shown in
Figure 2, the proposed scheme first initializes the number of flows to each cluster head. Each cluster head knows about the number of incoming and outgoing flows in their cluster. After the initialization of IntraSend scheduling, if the cluster has only incoming flows then the IntraSend scheduling can schedule the flows both in IntraSend and IntraRecv time slots. The cluster can utilize IntraSend time slots and cannot utilize IntraRecv time slots in-case if the IntraSend scheduling have both incoming and outgoing flows. For the InterComm scheduling the flows will be schedule from the source cluster head to destination cluster head. As the cluster head has high transmission power, therefore, it can utilize the InterComm time slots and cannot utilize the IntraSend and IntraRecv time slots. Finally, after the initialization of IntraRecv scheduling if the cluster has outgoing flows then IntraRecv scheduling can utilize both the IntraSend and IntraRecv time slots. Similarly, if the cluster have both incoming and outgoing flows then the IntraRecv scheduling can utilize the IntraRecv time slots and cannot utilize the IntraSend time slots.
We first schedule the flows from the source node to the destination node where each intra-cluster will have utilized only their respective time slots and cannot utilize each other empty time slots.
The cluster
H1 will schedule the transmission
ab and
bH1 of
F1 in the slot
#1 and slot #
5 respectively as shown in IntraSend Scheduling in
Table 1. Transmission
jc of
F5 cannot be scheduled in the slot #
1 due to the interference with the transmission
ab of high priority
F1. Therefore, the transmission
jc will schedule in the next time slots of IntraSend time slot #
5. For transmission
cH1 of
F5 as there is no interference with higher priority flow so it will be scheduled by
H1 in the slot #
9. Next cluster
H2 will scheduled transmission
ih of
F2,
pH2 of
F3 and
eg of
F4 in the slot #
1 due to no conflict with transmissions of high priority flows. Transmission
hH2 of
F2 will be scheduled in the slot #5 while transmission g
H2 will be cannot be schedule in the time slot#5 due to interference with transmission h
H2 of F2. Therefore, the transmission g
H2 of F4 will be scheduled in the time slot #9. Next in the InterComm scheduling, the flows from F1 to F5 are scheduled from the source node to the destination node as shown in the following
Table 2.
In the IntraRecv scheduling, each cluster will schedule the transmissions from the source cluster head to the destination node. For example, cluster H5 will schedule the transmissions of F1 and F5 while transmission of F2 and F4 will be schedule by cluster H3. H4 will schedule the transmissions of F3 from the source cluster head to their respective destination node as shown the
Table 3 below.
The proposed IntraSend scheduling system for cluster
H1 is shown in
Table 4. In
Figure 2, cluster
H1 has only outgoing flows, therefore, the cluster could use both
NIS and
NIR time slots for the IntraSend Scheduling.
F1 is scheduled first from node
a to
H1; thus
F5 is scheduled in
H1 cluster at time slot #
1. Since the transmission
ab of
F1 and transmission
jc of
F5 are conflicting with each other, the transmission
ab of
F1 is therefore scheduled in slot
#1 due to high priority. This is interesting, as there are only one incoming flow in cluster
H1. Thus, cluster H1 can utilize the IntraRecv time slots for IntraSend scheduling and schedule the transmission
jc of
F1 and
hH2 of
F5 in the same slot #
4 due to non-interference. Similarly, cluster H2 schedule transmission i
h of F2 in slot #
1. Transmission
pH2 of
F3 and
eg of
F4 are also scheduled in slot #1 due to non-interference with high priority flows. Cluster
H2 has also outgoing flows. In this, it can utilize the IntraRecv time slots. Therefore, transmission
hH2 is scheduled in time slot #
4. Transmission of F3 is already reached to cluster head so it cannot be schedule in slot #
4. Finally, the transmission
gH2 of
F4 is conflicting with transmission
hH2 of
F2, therefore, the transmission
hH2 is delayed by one slot and schedule in the next time slot #
5 as shown in
Table 4.
In InterComm scheduling, the global cluster header schedule the flows from
F1 to
F5 from source cluster head to destination head as shown in
Table 5.
IntraSend and IntraRecv scheduling can utilize each other slots but cannot utilize the InterComm time slots due to the high transmission power of the cluster heads. As shown in
Table 5, the global cluster head could schedule the flow
F1 from source cluster-head
H1 to neighbor the cluster head
H3. Furthermore, the neighbor cluster-head
H3 could schedule the data to destination cluster
H5. Similarly, for the
F2 flow, the global cluster-head cannot schedule the flow
H2H1 in the time slot #
6 due to the interference of high priority flow
F1. In the same way in slot #
7, the flow
F2 transmission
H2H1 is interference with transmission
H3H5 of
F1 and hence delayed. Flow
F2 transmission
H2H1 is scheduled in the time slot #
10 while transmission
H1H3 is scheduled in the time slot #
11. For flow
F3, the IntraSend transmission arrives at the source cluster-head in time slot #
1. Hence, the transmission
H2H4 and
H4H5 are scheduled by the global cluster-head in the InterComm time-slots #
2 and
3 respectively. For flow
F4, the transmission
H2H4 will be scheduled by the global head in the InterComm time-slot #
6, due to no interference with high priority flow. Hence, the transmission
H4H3 of
F4 has interference with transmission
H3H5 of high priority flow
F1 in the InterComm time slot #
7, therefore, the transmission
H4H3 will be delayed by one time-slot. Transmission
H4H3 of
F4 will scheduled in the InterComm time slot #
10 due to no interference with transmission
H2H1 of high priority flow
F2. For the flow
F5, InterComm scheduling the transmission
H1H2 is conflict with both transmission
H1H3 of
F1 and
H2H4 of
F4, therefore, transmission
H1H2 will be scheduled in the InterComm time slot #
7 as it has no interference with transmission
H3H5 of higher Flow 1. Similarly, transmission
H2H4 of
F5 cannot be schedule in the InterComm time slot #
10 due to interference with high priority flows transmission
H2H1 of flow
F2 and transmission
H4H3 of
F4. Finally, the transmission
H2H4 of
F2 will be scheduled in the InterComm time slot #
11, due to non-conflict with high flow.
Finally, consider the IntraRecv scheduling of clusters as shown in
Table 6. As shown in
Figure 2 cluster
H3,
H4 and
H5 have only incoming flows, so in these, IntraRecv scheduling does not differentiate IntraSend and IntraRecv time slots and can schedule IntraRecv flows both in IntraSend and IntraRecv time slots. For example, as shown in the
Table 6, cluster
H5 schedule
F1 transmission
H5m in the IntraRecv time slots while transmission
mq in the IntraSend time slots, as there is only incoming flows in cluster
H5. Similarly, cluster
H3 schedule transmission
H3k in the IntraRecv time slot whereas transmission
kl in the IntraSend time slots. Next cluster
H4 schedule transmission
H4r in the IntraRecv time time-slot while transmission
ro in the IntraSend time slots so utilized both the IntraSend and IntraRecv time slots as shown below in
Table 6.
To compare the previous and proposed system, we consider the flows arrived at their destination node. In the previous systems,
F1 arrived at destination node at time slot #
12 while in the proposed system
F1 arrived at destination node at time slot #
9. The reason is
F1 in the proposed system utilized the empty time slots #
4 of IntraRecv time slots while
F1 in the previous systems did not utilize the empty time slots of IntraRecv scheduling. F2 arrived at time slot #
16 to the destination node in the previous scheme while in the proposed scheme due to utilization of IntraRecv time slot #
4,
F2 reached to the destination node at slot #
13. In the previous scheme,
F3 arrived at the destination node in slot #
8 whereas flow
F3 in the proposed scheme arrived at the destination node at time slot #
5. Because
F3 in the proposed scheme utilized the empty IntraRecv time slot #
4, by the IntraSend scheduling, an empty time slot of IntraSend is utilized by IntraRecv scheduling. Similarly,
F4 in the proposed scheme arrived at the destination node at time slot#12 while in the previous scheme F4 arrived at time slot #
16 due to the non-utilization of the IntraRecv time slot. Finally,
F5 in the previous scheme arrived at the destination node at time slot #
24 while in the proposed scheme
F5 arrived at the destination node at time slot #
13 because the IntraSend scheduling utilized the empty time slots of IntraRecv. Similarly, the empty time slots of IntraSend is utilized by the IntraRecv scheduling. The flow diagram describing different states of the proposed scheme is shown in
Figure 3.