What are the differences between IET and UNH iSCSI implementation?
->" [This analysis] on open source iSCSI target software might interest you."
*UNH disk bypasses block-io and straightly use scsi functions
->pros:
may get extra performance gain from those skipped funciton calls. (not obvious in sequence io, but be better in random write)
->cons:
- skipping block-io also leave the benifits from using it, which provides raid, lvm ... devices access
the way UNH block io used make each access invoke disk access, but the chances of disk access could be reduced by using page cache which is used by Ardis. As a result in the network server benchmark tests, UNH disk is outperformd by others.
* the incompatibility problem of UNH package is on file-io mode, it seems the UNH file-io response without making sure the data has been writen to disks (do not provide "write durability")
* Ardis has a descendant, iSCSI Enterprise Target (IET)
* The analysis finished in 2004, that may not match the status of those iSCSI OSS today, after a years of development.
*find no relative proceeding research on google scholar search
* All the pictures used here come from the analysis.
