Neuro Cluster
The Neuro Cluster is a high-performance compute cluster located in the UC Berkeley data center.
It provides researchers with a reliable and secure platform for data analysis, data storage, and collaboration.
Hardware specifications
- Network storage servers provide over one petabyte of disk space in hardware RAID6 configuration to withstand simultaneous disk failures.
- Cisco 10Gbps network switches provide fast and reliable access to compute nodes and file servers.
- Compute nodes provide over 500 Intel CPU cores and 4TB RAM.
- Nvidia graphics cards, including Tesla V100, Tesla M2050, and GTX470, provide CUDA support and graphics acceleration.
Physical Security
The UC Berkeley Data Center is a hardened facility that provides a precision environment for the housing of mission-critical servers and related equipment.
The Neuro cluster hardware is secured in locked cabinets with dual power connections, backup UPS, and staffed Data Center Operations Desk.
Data integrity
- Data backups are stored for 7 days in a separate rack cabinet from the original data.
- Failed disk drives are permanantly erased via degaussing.
Computing support
The systems administrator is responsible for designing, configuring, and supporting the hardware and software aspects of the Neuro cluster. In particular, she directs hardware purchasing and maintenance of computing hardware and network devices. She installs, configures, and supports system software, scientific applications, and network services. She also provides detailed documentation about hardware and software resources, and assists researchers to run analyses more efficiently.