Version 3 (modified by rider, 14 years ago) (diff) |
---|
Deploy CUDA Cluster using DRBL
Hardware
CPU | Intel(R) Core(TM)2 Quad CPU Q6600 @ 2.40GHz |
Memory | 4GB/node |
Storage | 160GB/node |
Video Card | NVIDIA GeForce 9400GT 512MB/node |
Software
OS | Ubuntu 9.10 with Kernel_2.6.31-22 |
Driver | NVIDIA 256.40 |
CUDA | cudatoolkit_3.1 & gpucomputingsdk_3.1 |
Deployment keypoint
- CUDA library dynamic linker rebuild -> libcudart.so.3
- xorg.conf re-generate (using nvidia-xconfig)
- DRBL-CUDA: Fast GPU cluster deployment tool for CUDA
GPU Queueing & Management System
- Condor - queueing system
- Adding Graphics Card Support into Condor
- Hoopoe - Cloud Services for GPU Computing
- OpenPBS - workload management system
- Torque - resource management system