Whatever works best is fine by me; it doesn't have to be zipvfs. The documentation on CEROD is rather limited so it was a bit harder for me get a feel for how it works, what the tradeoffs might be, etc. One concern I had with CEROD was that it runs as a post processing step. That will probably provide better performance than on-the-fly compression since you know about all the data that's coming up front, but it will increase my peak disk capacity requirements further. An attractive feature of zipvfs (if I understood it correctly) is the plugability of compression algorithms. Being able to experiment with various algorithms (snappy, lz4, deflate, ...) to see what gives us the best results would be nice. I would happy to give both a try if that's possible so I can pick the most appropriate solution. I can go into more details on what I'm building if that helps, but that would have to be in a more private conversation. I don't think my employer would appreciate me sharing too many details just yet.