What I don't get however which feature of vmfs you exactly missing?
Being a cluster aware filesystem for starters. Everything else stems from that requirement.
All the supported volume manager / file systems for Proxmox are non-cluster aware, meaning they intrinsically assume the host has explicate and uncontested access to the hardware. This works great when it's local storage and is disastrously dangerous when it's non-local storage, aka Storage Area Networks. LVM is not cluster aware and neither is ZFS nor XFS or the other common file systems most linux people associate with. Proxmox tries to get around this by coordinating access to LVM volumes internally, cluster nodes talk to each other and say "resource X is using volume Y so don't anyone else attempt to use it".
This is how VMWare use's VMFS, and why they are the industry gold standard in this regard. Inside your SAN you slice off a LUN, which is a storage object that has it's blocks distributed across many disks and managed by the SAN's storage processors. That LUN is then mapped to the physical hosts that are going to use it over Fiber Channel or iSCSI. Every host then see's the LUN has a hard drive. One of the hosts will then format that hard drive with VMFS, which supports multiple hosts accessing it at once. Now every host in the cluster can read and write to the file system at once, sharing information without risk of them writing over each other or corrupting data. This is important because in VMWare your disks are just vmdk files on that shared LUN, vmdk which supports thin provisioning and fast snapshots. KVM's answer to this is QCOW2.
LVM/ZFS/Ceph, all of those storage solutions are for local storage, sometimes called Direct Attached Storage (DAS). Again this is because the Linux community grew up almost entirely in cost conscious environments with cheap hardware, and FC SAN's are expensive and generally existed with Unix like OS's like Solaris, AIX, HPUX, and later VMKernel/Photon OS. Linux would occasionally get a port but they didn't become serious until it started becoming popular as a replacement for those unix OS's.
https://en.wikipedia.org/wiki/GFS2
https://en.wikipedia.org/wiki/OCFS2
OCFS/OCFS2 came into existence because Oracle wanted to start shipping it's big, and incredibly expensive, RAC product on Linux instead of Solaris. RAC requires central shared storage that all the database servers use for their databases as any RAC cluster member can serve up any database. Solaris came with it's own cluster aware file system and so Oracle had to replicate that feature to support Linux, they even released it GNU GPL so as to be accepted into the Linux main kernel, which it was. It is a solution made specifically for compute clusters with shared SAN storage.
https://www.ibm.com/docs/pt-br/linux-on-systems?topic=ocfs2-history
Feature wise, it's supporting snapshots / thin provisioning on shared SAN storage. That is an extremely crucial component of VMWare central architecture and why VMWare become so popular. All the big players support some version of this.
Being a clustered filesystem for starters. All the supported volume manager / file systems for Proxmox are non-cluster aware, meaning they intrinsically assume the host has explicate and uncontested access to the hardware. This works great when it's local storage and is disastrously dangerous when it's non-local storage, aka Storage Area Networks. LVM is not cluster aware and neither is ZFS nor XFS or the other common file systems most linux people associate with. Proxmox tries to get around this by coordinating access to LVM volumes internally, cluster nodes talk to each other and say "resource X is using volume Y so don't anyone else attempt to use it".
This is how VMWare use's VMFS, and why they are the industry gold standard in this regard. Inside your SAN you slice off a LUN, which is a storage object that has it's blocks distributed across many disks and managed by the SAN's storage processors. That LUN is then mapped to the physical hosts that are going to use it over Fiber Channel or iSCSI. Every host then see's the LUN has a hard drive. One of the hosts will then format that hard drive with VMFS, which supports multiple hosts accessing it at once. Now every host in the cluster can read and write to the LUN at once, sharing information without risk of them writing over each other or corrupting data. This is important because in VMWare your disks are just vmdk files on that shared LUN, vmdk which supports thin provisioning and fast snapshots. KVM's answer to this is QCOW2.
LVM/ZFS/Ceph, all of those storage solutions are for local storage, sometimes called Direct Attached Storage (DAS). Again this is because the Linux community grew up almost entirely in cost conscious environments with cheap hardware, and FC SAN's are expensive and generally existed with Unix like OS's like Solaris, AIX, HPUX, and later VMKernel/Photon OS. Linux would occasionally get a port but they didn't become serious until it started becoming popular as a replacement for those Unix OS's.
https://en.wikipedia.org/wiki/GFS2
https://en.wikipedia.org/wiki/OCFS2
OCFS/OCFS2 came into existence because Oracle wanted to start shipping it's big, and incredibly expensive, RAC product on Enterprise Linux instead of Solaris. RAC requires central shared storage that all the database servers use for their databases as any RAC cluster member can serve up any database. Solaris came with it's own cluster aware file system and so Oracle had to replicate that feature to support Linux, they even released it GNU GPL so as to be accepted into the Linux main kernel, which it was. It is a solution made specifically for compute clusters with shared SAN storage.
https://www.ibm.com/docs/pt-br/linux-on-systems?topic=ocfs2-history
Feature wise, it's supporting snapshots / thin provisioning on shared SAN storage. That is an extremely crucial component of VMWare / Virtual Machine centric architecture and why VMWare become so popular. All the big players support some version of this. Your virtual machines and associated configuration are just files on shared disks and any host can pick them up and run them, meaning the individual host is unimportant and can be removed anytime for any reason without risk of data corruption. This is why we keep seeing people ask about it on various forums. That ability is very important for larger organizations that run traditional VM Centric infrastructure, you know the people Broadcom is screwing the hardest.
VMWare does VMFS6
Hyper-V does CSVS
Citrix/XenServer does GFS2 (though XCP-ng can't due to licensing restrictions)
The closest thing Proxmox has is NFS, which isn't something that can be done over FC though can be done over storage VLANs that would otherwise do iSCSI. It also requires the SP support it, which not all do and has some performance considerations as the SP has to assume control over all arbitration.
