This paper has more information on the implementation of PageRank with MapReduce, including its shortcomings, and optimizations that can be made to improve performance. MapReduce is often a bad fit for (iterative) graph algorithms simply because it was never designed with them in mind (e.g. the original paper never even discussed them). In the case of PageRank, using MapReduce involves passing a lot of data back and forth, which dominates runtime.
This comment was marked helpful 0 times.
ruoyul
Continuing the point above, since map-reduce uses a message passing model, data needs to be sent over the network between nodes, therefore if there were a lot of data to be sent, it will be costly.
This paper has more information on the implementation of PageRank with MapReduce, including its shortcomings, and optimizations that can be made to improve performance. MapReduce is often a bad fit for (iterative) graph algorithms simply because it was never designed with them in mind (e.g. the original paper never even discussed them). In the case of PageRank, using MapReduce involves passing a lot of data back and forth, which dominates runtime.
This comment was marked helpful 0 times.
Continuing the point above, since map-reduce uses a message passing model, data needs to be sent over the network between nodes, therefore if there were a lot of data to be sent, it will be costly.
This comment was marked helpful 0 times.