Quantum Forum for DXi V5000
Trying to understand deduplication logic in V1000
- If multiple CIFS shares are on the DXi are those shares islands of deduplication or is the deduplication global?
- Same question for OST but there you could have multiple LSU’s under 1 storage server. Is each storage server an island of dedupe and the LSU’s data is comingled?
- If that is true, an MSP could “segregate” clients data by creating multiple storage servers?
Or is all data (CIFS) (OST) globally deduped?
Replies to This Discussion
-
Permalink Reply by Mark Thacker on
-
On all DXi systems, all data, regardless of the presentation layer, is put into a common deduplication blockpool.
So, we do not isolate deduplication on a per share or per presentation layer basis.
If you ingest say a backup set via CIFS and then also ingest that same set via NFS for example (or OST), the second set of data would essentially be 100% duplicate data and thus only the indexes to the data set would be updated, no additional data would land in the blockpool.
For multi-tenant environments, we find that it's access to the data, not the intermingling within the blockpool, that a lot of customers want. Each share or LSU on a DXi can have it's own userid/password for differentiated access, but would still be using one common Windows Workgroup, Windows Active Directory Domain or common set of LSU userid/passwords.
Because the V1000 is a virtual machine using thin provisioning, we do have customers and service providers that stand up a dedicated DXi V1000 per customer with no intermingling of any data or authentication. The thin provisioning of VMware, plus the access controls allowed by VLANs provides separation of data.
Remember, if you want to separate data, then you need separate storage pools for dedup. If you want separate access controls, there is some of that available today.
-
Permalink Reply by bob waters on
-
Perfect!
Thanks
© 2024 Created by Quantum Forum V.
Powered by
