Lobster: Load Balance-Aware I/O for Distributed DNN Training.

Jie Liu Bogdan Nicolae Dong Li

Published in: ICPP (2022)

Keyphrases

load balance
load balancing
training process
computing resources
round robin
disk array
scheduling algorithm
inter processor communication
distributed systems
multiple disks
fault tolerant
database
parallel machines
file system
end to end
power consumption
response time