There's a few problems here.
Your ObjectID contains KAA. That means you have 5+ billion files in a single Application Group. IBM CMOD likes economies of scale -- loading thousands of documents at once. You're also losing the benefits of CMOD's compression algorithms as well -- because they perform best when the input files are enormous.
Second, if you're loading individual documents at a time into an Application Group, and the expiration type is set to 'Load', and you're doing LOTS of concurrent loads, you're going to have contention issues on the IBM CMOD arsload table as each load tries to insert a row into the table.
Third, you appear to be using TSM - so there's a possibility you might be using 'Migrate on Load'. While this is normally a good idea (for data redundancy), it slows down loading. You can change the AG config to move the data 'At next cache migration' -- but you *absolutely* must use the IBM CMOD arsmaint utility to migrate data -- preferably several times a day.
Before doing anything else, I'd consider moving historical data (a year at a time) into a new Application Group with a better configuration, then binding them together at the folder level.
Of course, this is just one Application Group -- if your others are similarly used (and misconfigured) then you probably have other App Groups to fix as well before performance improves.
-JD.