On 2020-02-13 10:26, Janne Johansson wrote: > In the larger scheme of things, "one smb server", "one nfs server" or "one > cephfs MDS" doesn't differ much. > > You need some kind of box to translate from object storage (regardless of > if this is iscsi, ceph or something else) to a kind of filesystem that can > give you some extra guarantees (like being a POSIX filesystem) and keep > state of locks and atomic ops so that more than one client-or-process can > access the shared space without stepping on each others toes. Indeed. Some of these protocols may have higher overhead when it comes to translating the data through multiple storage protocols into Ceph's RADOS objects. The fewer "layers", the better the throughput/latency. > Since there aren't (as far as I know) any native ceph access methods for > windows, even a single windows client still need someone somewhere to > translate the storage into a form usable by it. Not yet, but this is being worked on: https://github.com/ceph/ceph/pull/31981 https://github.com/petrutlucian94/ceph/blob/windows/README.windows.rst The intention is to implement a native windows client that will eventually be able to consume both RBD and CephFS resources. Lenz -- SUSE Software Solutions Germany GmbH - Maxfeldstr. 5 - 90409 Nuernberg GF: Felix Imendörffer, HRB 36809 (AG Nürnberg)
Attachment:
signature.asc
Description: OpenPGP digital signature
_______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
