single write/multiple read, virtual dataset, scalable chunk indexing, free file space tracking, collective metadata I/O