Hi guys,
I'll have a future Ceph deployment with the following setup :
- 7 powerful nodes running Ceph 15.2.x with mon, rgw and osd daemons
colocated
- 100+ SATA drives with EC 4+2
- every OSD will have a large NVME partition (300GB) for rocksdb
- the storage will be dedicated for rgw traffic using Swift (no cephfs,
no rbd)
- we will probably have a lot of 4MB ceph objects (more than 400k
millions) after the first year.
Does it still matter to have the rgw index pool on dedicated SSD/NVME
drives or it's good enough to spread it on many SATA drives with 3x
replication and a large PG number ?
Thanks.
_______________________________________________
ceph-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]
