Quick analysis of the sample 1M emails have shown that:

• 97% of mails are less than 128K
• 95% of mails are less than 64K

This figure would vary depending on use case (e.g. free webmail vs corporate email), but in most cases we would expect >90% emails to be less than 128K.

Taking into account that these messages will be compressed with 2:1 ratio, it should be convenient to store 32K-64K blobs in Cassandra.

Internal implementation of Blobs can be designed after CassandraFS to provide good scalability.

For the beginning we can use single block (row) and single sub-block (column) implementation where max block (and sub-block) size is 128K. It should possible to add multi-block support in future.

Threshold between internal (C*) and external (Cloud) blob storage should be configurable depending on blob size.