Brief list of recomandations extracted from Ceph official site:
- Run OSD data and OSD journals on separate drives
- Do not run multiple OSDs on a single disk
- Do not run OSD and a monitor or a metadata server on a single disk
- Metadata servers and monitors should have plenty of RAM (1GB of RAM per daemon instance)
- Metadata servers should have significant processing power (e.g., quad core or better CPUs)
- Run less performant CPU for Monitors as are not CPU intensive
- BTRFS is not quite stable enough for production
- BTRFS has the ability to journal and write data simultaneously, whereas XFS does not.
- Run OSD with larger drives as better price per gigabyte
- Run a dedicated drive for the operating system and software
- Run one drive for each Ceph OSD Daemon
- iSCSI gateway can be installed on CentOS or RHEL only (July 2020)
- Use of SSDs to improve performance
- OSDs may see a significant performance improvement by storing an OSD’s journal on an SSD
- Each host have at least two 1Gbps network interface controllers (NICs)
- Consider starting with a 10Gbps network in your racks