Table of contents

Schema for compute node rack rack #23

height unit usage OS eth0 eth1 eth2 eth3
42 sysgen GPU01 Intel based dual Tesla C1060 Debian (x86_64)        
41
40
39 sysgen GPU02 Intel based dual Tesla C1060 Debian (x86_64)        
38
37
36 sysgen GPU03 Intel based dual Tesla C1060 Debian (x86_64)        
35
34
33 sysgen GPU04 Intel based GTX295 Debian (x86_64) / XP (x86_64) / Vista (x86_64)      
32
31
30
29 clustervision GPU05 AMD based dual Tesla C1060 Debian (x86_64)        
28
27
26
25 clustervision GPU06 AMD based dual Tesla C1060 Debian (x86_64)        
24
23
22
21 clustervision GPU07 AMD based dual Tesla C1060 Debian (x86_64)        
20
19
18
17 clustervision GPU08 AMD based dual GTX285 Debian (x86_64) / XP (x86) / Vista (x86)        
16
15
14
13  
12 "Titan01" post-processing box 64GB
11 "Titan02" post-processing box 64GB
10  
09 Orion
08
07
06
05
04 s13
03
02
01

Windows installations (GPU04 / GPU08)

Initial software packages

  • BOINC
  • Chipset Drivers
    • GPU04: Intel (Asus P5E Motherboard)
    • GPU08: NVIDIA (Asus M3E-HT Deluxe Motherboard)
  • CopSSH (freeware / open source)
  • NVIDIA Graphics Driver
  • NVIDIA CUDA 2.2 Toolkit and SDK
  • Sophos Anti Virus (RRZN)

Initial configuration

  • Main boot-manager: grub (installed in MBR)
  • Using Windows Vista boot-manager for XP and Vista (launched by grub)
  • Partitioning: XP (20 GB), Vista (40 GB), Windows 7 (40 GB)
  • Remote desktop access enabled
  • Firewall disabled
  • Automatic update notifications only (download/install on demand)
  • Dedicated single account: "root" (ssh access enabled)

Additional notes

  • KVM-Switch for users: kvm-gpu1:
    • 1 session / 4 participants
    • User name: user
    • Password: please ask admins (atlas-admins@aei.mpg.de)
    • Availability: internal institute network (10.117.0.0)
  • KVM-Switch for admins: kvm-gpu2:
    • 1 session / 4 participants
    • Availability: internal institute network (10.117.0.0)
    • Use OSD (CTRL-ALT+M) to select host

  • change kernel driver version:
    • right now: sudo /local/cuda/switch_nvidia_kerneldriver.sh (temporary hack)

This topic: ATLAS > RackSchemes > ComputeNodeRack23
Topic revision: 08 Sep 2009, CarstenAulbert
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Foswiki? Send feedback