Can't find StarWind Server by scanning

[anton (staff)]

No. What you're talking about is COW (Copy-on-Write) and 1) it's NOT file system cache 2) most of the software vendors prefer to do it on the server side. As a result you'll have BASIC image which is the same for all the initiators and "delta" images all different. With this scenario you can save on disk space (poor man's de-duplication) but you CANNOT share the written content, only BASIC data already on the BASIC image.

Looking at that link you suggested, does the term "distributed caches" imply that in iSCSI connection each Initiator saves locally a copy of data that was ever written to each Target during current session? Once the session is closed, that local file copy is then discarded at local disk being dismounted? If that's correct, where to find that copy before the session is closed? Does the local disk mounted on a free space of an existing locally Volume, or in unformatted space on a local drive if such space exists?

[Zamar25]

Then what does the term "distributed caches" mean in practice? Does it mean that each connected to SAN volume's file system holds in RAM its current file table copy, and an arbitrator like metaSAN strives to update these copies in real time to keep them exactly the same?

[anton (staff)]

Cluster file systems and arbiters for generic FSes work in the different way. File systems DO keep caches coherent (pretty much what StarWind does for HA nodes) and arbiter makes generic file systems flush portions on volume (read: disables cache from time-to-time). That's why monolithic FS performance is expected to be better. At least in theory.

Then what does the term "distributed caches" mean in practice? Does it mean that each connected to SAN volume's file system holds in RAM its current file table copy, and an arbitrator like metaSAN strives to update these copies in real time to keep them exactly the same?