The invention discloses a rainbow table parallel system design method and device based on MapReduce. The design method includes a rainbow table data generation step and a rainbow table decryption step; the method includes aiming to each rainbow chain of a rainbow table, allowing a map module to acquire a chain head node data vector and chain end node data vector of the rainbow chain, recording the chain head node data vector as (img file = 'DDA0000586158180000011.TIF' wi = '88' he = '72' /), recording the chain end node data vector as (img file = 'DDA0000586158180000012.TIF' wi = '95' he = '66' /), adopting EPi as Key, adopting SPi as Value, and transmitting <EPi, SPi> data to a Reduce module of the rainbow chain; allowing the Reduce module of the rainbow chain to reorganize the <EPi, SPi> data, acquiring a data structure that (img file = 'DDA0000586158180000013 .TIF 'wi =' 400 'he =' 72 '/), presenting a chain table as (img file =' DDA0000586158180000014.TIF 'wi =' 238 'he =' 68 '/), and allowing the value of the chain table to be a set of chain head node data vectors of the rainbow table with all chain end node data vectors serving as the EPi; inserting all reorganized rainbow table data and fresh & lt; Key, Value & gt; data into a database NoSQL; analyzing Hash H0 to be decrypted, and mapping the H0 as nodes Nodei of the rainbow chain; converting the Nodek as the chain end node data vectors EP'i; adopting the EP'i as the key to read the reorganized rainbow chain data from the database NoSQL, and analyzing and acquiring the H0 passwords.