Korean Android Trace/Measurement

They build a FTL simulation tool, and also back-mapping from block number to inode number to file path.

Using that we ware able to measure:

1. Write size (sectors) and write requests distributions on file types and application.
    For almost every applicton:
    -- ext4 journal writes are the most
    -- then android and database temporal file I think

2. Avg overwrites on a block distribution on file types and application
   for each application:
     -- many overwrites on the same block for ext4 journal files and database/database journal files

3. Avg write size per request based on file types and application
    for each application:
     -- 1.x blocks per write request for almost all kinds of writes
     -- 6-7 blocks per write for ext4 journal

4. Traffic source --- how much data is issued for write for each type of file (per application)
     -- for every 1kb user data, seems like 3kb of android&kernel temporal data is written, and 2kb FTL garbage collection & FTL other writes