Evan Zhou

@manishrjain The most important optimization is the SST format change and table iterator, merge iterator. The scan performance has improved multiple times, even faster than RocksDB.

The initial fork picked this optimization https://github.com/dgraph-io/badger/pull/481 And then those are the related PRs, the earlier ones are more important. https://github.com/coocood/badger/pulls?utf8=%E2%9C%93&q=is%3Apr+is%3Aclosed+optimize

No in the near future.

@Connor1996 Please try and check if this https://github.com/coocood/badger/pull/171 change fixes this issue.

@tiancaiamao PTAL

@gregwebs I have another keyspace design for multi-tenant at https://github.com/ngaut/unistore/pull/393 The key difference is the cluster id is used as tenant ID, and the prefix for tenant id is fixed.

LGTM

> After hacking at this for a bit I am now questioning whether this is worth the effort. There are a lot of places where prefixes could now be used....

Not necessary, if the cache keep running and write data at 1GB/s, it will take more than a hundred years to overflow.

For persistent file format, we better add file format version in case we change it later, and calculate the checksum of the content.