查看文章

Conventional distributed systems manage a cluster of computing nodes through cluster-wide coordination with respect to communication, computation and storage, represented by Hadoop Clusters and Spark Clusters. Huge data can be partitioned and distributed by partitions to different nodes in a cluster. Computation can be done in either local mode or distributed mode. In local mode, computation needs to handle both location computing and data movements to and from other nodes in the cluster. In distributed mode, the local computation needs to be synchronized through inter-node communications across the cluster. For huge data movements across a compute cluster, the inter-node communication for distribution synchronization can be prohibitively expensive.

In the age of big data powered Artificial Intelligence (AI) and Machine Learning (ML), Data has become the No. 1 in exponential growth, faster than big …