There are two major products that came out of Berkeley: LSD and UNIX. We don’t believe this to be a coincidence.
—Jeremy S. Anderson
Like LSD and Unix, Spark was originally conceived in 20091 at Berkeley,2 in the same Algorithms, Machines, and People (AMP) Lab that gave the world RAID, Mesos, RISC, and several Hadoop enhancements. It was initially pitched to the academic community as a distributed framework built from the ground up atop the Mesos cross-platform scheduler (then called Nexus). Spark can be thought of as an in-memory variant of Hadoop, with the following key differences:
Directed acyclic graph