Skip to main content
. Author manuscript; available in PMC: 2017 Feb 10.
Published in final edited form as: SIAM J Sci Comput. 2016 Jun 1;38(3):C179–C202. doi: 10.1137/15M1014784

Table 1.

Comparison of storage types available to MOLNs distributed workflows.

Type Advantages Disadvantages
SharedStorage No additional cost for read/write
Fastest throughput for small clusters
No management of remote data		 Total storage limited to Controller disk size
Nonredundant storage
Throughput limited on large clusters
PersistentStorage Persistent data
Designed for extreme scalability		 Storage and access incur cost
LocalStorage Best data locality
High I/O throughput		 Nonrobust to worker failure
Increased complexity for developer
No-Storage Best parallel scaling
No cost for data storage		 Data must be recomputed for analysis