C3DDB Compute, Storage, and Networking Resources
The C3DDB compute cluster currently consists of 133 compute nodes with a total of 7200 cores 61 terabytes of main memory, an FDR infiniband interconnect, a 2 Pebibyte (2.24 Petabyte) Lustre file system, and a 54 Terabyte NAS.
- 100 compute nodes, each consisting of a Dell PowerEdge R715 chassis containing:
Four 16-Core AMD Abu Dhabi Opteron processors (AMD 6376) running at 2.3Ghz
256GB main memory
500GB scratch storage
- 32 compute nodes, each consisting of a an HP DL580 chassis containing:
Two 10-Core / 20 Thread Intel Ivy Bridge processors (Intel E7-4830v2) running at 2.2Ghz
1 Terabyte main memory
3 X 3 x nvidia K40c GPU accelerators
- 12 compute nodes, each consisting of a Dell PowerEdge C4130 chassis containing:
Two 14-Core / 28 Thread Intel Broadwell-EP processors (Intel E5-2680v4) running at 2.4Ghz
128GB main memory
1000GB scratch storage
- 42 Intel 7000 chassis, each containing four Intel KNL 7210 compute nodes
168 KNL 7210 nodes total
- 1 large-memory compute node consisting of an SGI UV2000 containing:
Twenty 8-core Intel Ivy Bridge Processors (Intel E5-4650 v2) running at 2.4GHz
Four Terabytes of main memory
- Master and gateway nodes to support login, management, and network access
- License server
- All of the nodes in the system run RHEL 6.5, and are managed by a SLURM resource manager.
- 2.24 Petabytes (2 Pebibytes) of storage for scratch and project directories, managed by a high performance LUSTRE file system.
- 54 Terabytes of NAS storage for home directories, supported by snapshot and backup services.
- dbGap-compliant directories for sensitive data.
- For ordinary use:
- The C3DDB is accessible via SSH, SFTP, and SCP from anywhere on the public internet.
- For high speed data transfers:
- All computing systems at the MGHPCC, including the C3DDB, are accessible via 10Gbps links to the five MGHPCC member universities.
- The C3DDB is also accessible via the Northern Crossroads node at the MGHPCC.
- A Globus Connect Data Transfer Node is also available for those who need a highly optimized data transfer path.