Question-21: When should we delete the data from commit log or commit log file itself?

Answer: Once all the data from the Memtable Is flushed in the SSTable than you can archive it or delete the commit log files.  

Question-22: What information is stored by the gossip?

Answer: Gossip is a peer to peer communication protocol which is used to discover and share location and state information about the other nodes in the cluster. Gossip information is persistent locally by each node to use immediately when a node restart.

Question-23: How do you define a partitioner?

Answer: It is the responsibility of the partitioner to distribute data evenly across the nodes in the cluster, and also helps in load balancing. Partitioner determines which node in the cluster should receive the first replica of the data and how to distribute other replicas across the other nodes in the cluster. Similar to other data bases each row of the data is uniquely identified by a primary key. And partitioner derive the hash value from the primary key of a row, which works as a token and based on the token value partitioner found the nodes in the cluster to place the replica of that row. 

Question-24: Which of the available partitioner should be good choice for most of the cases?

Answer: The Murmur3Partitioner is the default partitioning strategy in most of the cases.

Question-25: How does Cassandra determines on which node the first replica of the data should be placed? 

Answer: There is a replication strategy which determines on which node the first replica and another replica should be placed. The NetworkTopologyStrategy, is used in most of the cases, because for most of the deployment it is easier to expand to multiple datacenters when this strategy is used.