==============

Must provide detailed answers to get upvote

Thanks

(Parallel and distributed systems) Suppose you want to parallelize a program on an 8-node cluster. Assume 20% of the program is purely serial code that cannot be parallelized, what is the speedup you can achieve by perfectly parallelizing this program on the 8-node cluster compare to run the same code on one node? (Hint: Use the Amdahl's law.)