Each zone is constructed as virtual hierarchical that are geographic -specific. Each zone has a pre-defined number of nodes and a coordinator 〖C(i)〗_11^1to these nodes. The coordinator is assigned based on the proposed IHRW hashing algorithm, as illustrated in heuristic 3.
Heuristic 3: The adaptive HRW protocol proposed in figure 5 is implemented in each level of the virtual hierarchical. Suppose the system contains n_od nodes that will be divided into number of Rendezvous geographic Zones (〖RG〗_Z) based on their geographic area. Each zone will have number of nodes n_od ranges from max┬iceli to min┬iflo depend on their area load. This will achieves O(log(AVG [max┬iceli , min┬iflo] ) running time.
These zones are categorized as follows and so on. Then instead of hashing all the nodes in the system, we start hashing each node in a specific zone by using spooky hash function illustrated in next subsection, then based on the hashes obtained the nodes are organized in levels (level 1 to 3 based on number of nodes) and the node with the highest weighted hash in the upper level will be assigned as coordinator of this level 〖C(j)〗_(m.x)^n, where n is the zone number, m is the level the coordinator in and x is the coordinator ID. Then the coordinators nodes is hashed and the highest weighted hash in them will be …show more content…
2014; Ramakrishnan et al. 2015]. Workload C of YCSB is used in load mode to upload the data to one cluster of 8 nodes. Figure 12 shows the performance of Cassandra with ByteOrder partitioner, the load is condensed in a one or two nodes which result in uneven distribution of data. Figure 13 demonstrates that using Murmur (the current default Partitioners) in Cassandra the load is almost distributed on the 8 nodes but not
Heuristic 3: The adaptive HRW protocol proposed in figure 5 is implemented in each level of the virtual hierarchical. Suppose the system contains n_od nodes that will be divided into number of Rendezvous geographic Zones (〖RG〗_Z) based on their geographic area. Each zone will have number of nodes n_od ranges from max┬iceli to min┬iflo depend on their area load. This will achieves O(log(AVG [max┬iceli , min┬iflo] ) running time.
These zones are categorized as follows and so on. Then instead of hashing all the nodes in the system, we start hashing each node in a specific zone by using spooky hash function illustrated in next subsection, then based on the hashes obtained the nodes are organized in levels (level 1 to 3 based on number of nodes) and the node with the highest weighted hash in the upper level will be assigned as coordinator of this level 〖C(j)〗_(m.x)^n, where n is the zone number, m is the level the coordinator in and x is the coordinator ID. Then the coordinators nodes is hashed and the highest weighted hash in them will be …show more content…
2014; Ramakrishnan et al. 2015]. Workload C of YCSB is used in load mode to upload the data to one cluster of 8 nodes. Figure 12 shows the performance of Cassandra with ByteOrder partitioner, the load is condensed in a one or two nodes which result in uneven distribution of data. Figure 13 demonstrates that using Murmur (the current default Partitioners) in Cassandra the load is almost distributed on the 8 nodes but not