EDFS: a semi-centralized efficient distributed file system

In this paper we present (1) EDFS, a simple and semi-centralized efficient distributed file system (for cloud computing environments) and (2) a load balancing and resource sharing scheme for EDFS. EDFS uses a central front-end server only to distribute keys (hashed filenames or indices) to Namenodes. Each Namenode is then responsible for holding the corresponding file (directory) information such as the i-node (metadata) and pointers to where each block of the file is stored in the cloud. Each node is also responsible for every operation (open, close, delete, modify) of the files (directories) whose keys it has. This will enable EDFS to distribute the cloud computing to all nodes in the cloud unlike the current Hadoop File System (HDFS) and Google File System (GFS) which load the Namenode with almost every operation of every data block, potentially making the Namenode a resource bottleneck.