MapleJuice

A Hadoop Map-Reduce like framework implemented using Golang; A round 3 times faster than Hadoop MapReduce (by issuing 5 nodes for each application). The system is supported by a simple distributed file system and a gossip style based failure detector.

Setup

cd /MP3
sh build.sh
at the master node, run sh vote_build.sh if you need input data for voting sets application, otherwise the input data is for word frequency
then run:
./maplejuice on every single node

Maple

in the system, input maple
then
<maple_exe> <num_maples> <sdfs_intermediate_filename_prefix> <sdfs_src_directory>

Juice

in the system, input juice
then
<juice_exe> <num_juices> <sdfs_intermediate_filename_prefix> <sdfs_dest_filename> if the maple function executed before this juice function is an identity function, please directly run:
<juice_exe> <num_juices> <sdfs_intermediate_filename_prefix> <sdfs_dest_filename_of_last_juice>

Other commands:

`You can use the following commands:

Failure detector related

• join: make this machine join the membership list
• leave: leave from the membership list
• change: change from GOSSIP to ALL_TO_ALL, vice versa
• ls: show all machines and their statuses in the membership
• id: show ID of the current machine
• mode: show current failure detecting mode

File system related

• put: insert/update a file to the distributed file system
• get: get a file from the SDFS
• delete: delete a file from the SDFS
• store: show the files stored on the current node
• ls: show the positions that a particular file is stored at