▸ Primary use case: distributed key-value store for blobs ▸ Existing distributed filesystems have too much overhead for metadata and support for all kinds I/O operations ▸ Key value stores are not optimized for blobs, ie. support zero-copy reads, streaming, etc.