RoCE (RDMA over Converged Ethernet) and Persistent Memory are features that are supported by Oracle Linux. The minimum version of Oracle Linux required to support RoCE and Persistent Memory is Oracle Linux 7.9.
RoCE is a low-latency, high-bandwidth networking technology that enables fast data transfer between servers using converged Ethernet networks. It is supported by Oracle Linux 7.9 and later versions, which includes the necessary kernel modules and user-space libraries to enable RoCE.
Persistent Memory is a technology that allows data to be stored in non-volatile memory (NVM) devices, such as NVDIMMs (Non-Volatile Dual In-line Memory Modules), which can be accessed at near-DRAM speeds. Oracle Linux 7.9 and later versions include support for Persistent Memory, which includes kernel modules and user-space libraries to enable Persistent Memory.

The steps and the order in which they should be executed to create an ACFS Filesystem for use as a staging area for data loads are:
asmcrad volcreate -G data -s 1024G acfs_dataload_vol : This command creates a new volume in the data diskgroup, with a size of 1024 GB, and named acfs_dataload_vol.
asmcmd vol info -G data acfs_dataload_vol | grep Device: This command will show the device name of the newly created volume
/usr/sbin/mkfs -t acfs /dev/asm/ac:fs_dataload_vol-123: This command creates a new ACFS filesystem on the volume created in step 1 sudo /u01/app/19.0.0.0/grid/bin/srvctl start filesystem -device /dev/asn/acfs_dataload_vol-123: This command starts the filesystem created in step 3 sudo chovn -R oracle:dba /mnt/acfs_dataload_vol: This command changes the ownership of the /mnt/acfs_dataload_vol to the oracle:dba user.
/sbin/acfsutil snap create -v data_load /mnt/acfs_dataload_vol: This command creates a snapshot of the filesystem to be used as a staging area.
acfsutil size +2S0G -d /dev/asm/acfs_dataload_vol-123 /mnt/acfs_dataload_vol/ : This command increases the size of the filesystem by 2GB
/u01/app/19-G.O-Q7grid/bin/srvctl add filesystem -device /dev/asm/acfs_dataload_vol-

Persistent Memory Commit Accelerator reduces redo log write latency by using Persistent Memory and RDMA before flushing to Flash then disk. This helps to further reduce redo log write latency by utilizing the speed and low latency of Persistent Memory, along with the Remote Direct Memory Access (RDMA) protocol, to commit changes to disk faster. This allows the system to quickly commit changes to disk, resulting in improved performance and reduced latency.