Question

If pagerank computation takes 10 iterations to converge, how many rounds of map reduce you need to compute it in parallel? Also mention one performance drawback of using mapreduce to compute page rank?

Answer 1

To compute PageRank in parallel using MapReduce, you need 10 rounds of map reduce. One performance drawback of using MapReduce is the high communication overhead.

Step-by-step explanation:

To compute PageRank in parallel using MapReduce, you would need to perform 10 rounds of map reduce, each round corresponding to one iteration of the PageRank computation. For example, in each round, the map function will process the adjacency list of a web page to calculate the contributions from its outgoing links, and the reduce function will aggregate these contributions to update the PageRank values.

One performance drawback of using MapReduce to compute PageRank is the high communication overhead. MapReduce involves shuffling and sorting of intermediate key-value pairs, which can create a significant amount of network traffic. This can slow down the computation, especially when the graph is large and the amount of intermediate data is substantial.