Why is Ceph and its CRUSH algorithm less used for big data analytics?

Question

As Ceph and its CRUSH algorithm ruled out the issue of metadata server's contention, and surely decreased object fetching latency by removing the RPC to query object location, why it is less adopted for big data analytics? Ceph's popularity is not comparable to HDFS, even though what HDFS can do (cluster member changes, replication rule sets, locality of compute and data), can also be done by Ceph.