Aurora scales DAOS to 800 servers

The Argonne Leadership Computing Facility (ALCF) has finished expanding the Distributed Asynchronous Object Storage (DAOS) system on Aurora, its Intel-HPE exascale supercomputer. The system started with 128 storage servers, grew to 370 in an intermediate phase, and now runs in its final configuration of 800.

ALCF benchmarked the full system at up to 30 terabytes per second using 2,048, 4,096 and 8,192 Aurora compute nodes. According to ALCF, the results show that performance keeps scaling and does not collapse at extreme client counts. Aurora, which integrates DAOS directly with its compute fabric to support I/O and checkpointing, also remains No. 1 on the ISC26 IO500 Production list, based on earlier results from 2023.

The DAOS Foundation congratulates ALCF, a founding member, on the milestone. Read the original announcement on the ALCF website.