Hardware
Storage System
The DORS primary disk consists of a DataDirect Networks (DDN) SFA-12K40 storage system, located in the Hill Data Center. Highlights include:
- Dual storage controllers with active-active failover eliminates downtime during routine maintenance
- Filesystem metadata is stored on Solid-State Disk (SSD), minimizing latency while browsing files
- 427 terabytes of usable capacity in 350 disk drives
- 490 empty disk drive slots, enabling low-cost expansion
- 80x 6Gbps SAN fabric connections and a QDR Infiniband network provide up to 40 Gigabytes per second of throughput to clients
Tape Backup System
The DORS backup system consists of an IBM tape library with the Advanced Library Management System and Tivoli Storage Manager (TSM) software. This system is located in the VUH datacenter, providing additional data security in case of building/infrastructure failure in the Hill Data Center (and vice-versa). Key capabilities:
- Logical partitioning enables multiple tape services to simultaneously share the robotics
- 6x LTO-6 tape drives (2.5TB capacity per cartridge)
- >2300 slots for up to ~6 petabytes of shared capacity
- 16Gbps Fibre Channel fabric for maximum throughput
- 20TB disk cache minimizes nightly backup window
Access and Connectivity
DORS runs on IBM’s General Parallel Filesystem (GPFS), providing high-performance, parallelized throughput to ACCRE via native GPFS connections to all gateways and compute nodes. Access to systems outside of ACCRE is supported by Clustered NFS (CNFS) and Windows file sharing (CIFS). GPFS, CNFS, and CIFS are all configured with multiple redundant servers, providing automated transparent failover in case one server requires maintenance, or for carrying out rolling upgrades of system software with no downtime. Network connectivity consists of multiple redundant 10Gbps Ethernet links from every DORS server to ACCRE (40Gbps total), main campus (20Gbps total), and VUMC (20Gbps total).