Problem

• I/O is the bottleneck of a system
• Read is easy to deal with, but write operations could not scale-out due to multiple replication

Solutions

• use memory to store most recent files
• don’t do file replications
  • it would do checkpoints of all files
    • needed information is lineage: input file, binary programs, execution configuration, dependency type
    • discard old file versions after checkpoint
    • if file is too big to fit in memory, checkpoint leafs firstly during operation burst, then checkpoint the rest of non-leaf files
    • once one file have checkpoint, I could discard it from memory
  • but once the storage is down forever, it could not be recovered without previous backups or replications