Andersen et al. [9] proposed the FAWN system for datacenter key-value stores. They suggest a scheme that builds
on low-power cores and Flash storage to construct an efficient high-performance key-value storage system. The proposal includes hardware-software co-design and their software innovations are primarily aimed at improving Flash
storage’s lifetime. The basic premise of their system is
to match compute and I/O abilities for a specialized task
of key-value datastore. Our system however is suited to
general computational requirements of modern datacenters,
while FAWN is targeted at a specialized application. The
main ideas in our work are orthogonal to those in the FAWN
architecture.