When a client writes a file to the HDI file system for the first time, it is kept resident on HDI until the next replication event, when HDI replicates all new files to HCP. The original default for replication from HDI to HCP was once per day, but the can be adjusted down to the level of minutes. From the HDI 6.1.0 release the default migration interval was reduced to once per hour. The migration interval however can be set in a policy and can be made to replicate the data from HDI to HCP as frequently as every 15 minutes.
Once data is replicated from HDI to HCP, the URL of the data on HCP is added to the metadata of the file on HDI and the file is identified as a candidate for stubbing
Stubbing is a different operation from replication and operates independently from replication. A file on HDI will only be stubbed if it has been marked as a candidate for stubbing and the URL of the data on HCP has been added to the files metadata on HDI.
By default, when the file system capacity reaches 90%, HDI determines which files are eligible for stubbing based on a psuedo Least Recently Used (LRU) algorithm. The LRU process groups the files by access time or create time (atime or ctime). Only files that have already been replicated to HCP can be considered for stubbing. The oldest groups are stubbed first. Stubbing will continue until the file system used capacity drops below 90% again. This stubbing activity happens every hour.