PARALLEL MONSTER:
A 64-Opteron NSF-sponsored Linux Cluster at MSU
Assembled: December 2003
Click a picture to enlarge; for more (historic) pictures click HERE.
Budget/Sponsor
- $100,000
- National Science Foundation
- Grant #: CTS-0319555 (PI: Dr. R. M. Zaritski; MRI, 9/1/03-9/1/06)
Performance
- Peak: 180 Gigaflops
- Linpack (HPL) benchmark: 60 GFlops
(on a 40,000x40,000 matrix, with LAM/ATLAS)
Nodes
- Dual AMD Opteron 240 (64bit, 1.4GHz)
- 31 Slaves, 1 Master
- Memory: 2GB each Slave, 4GB Master (DDR SDRAM 333MHz ECC Reg)
- Disks: 80GB IDE HD (2x146GB SCSI on Master)
- Microway NodeWatch cluster management hardware
Node Interconnect
- Switched Gigabit Ethernet over Cat 5e/6 Copper cables (1000BaseT)
- HP Procurve 4108GL 36-port GigE Switch
Cabinet and Power
- Microway 44U rackmount cabinet (with 3 door fans)
- Six 120V 15A power lines
- APC Smart-UPS power backup (750VA, 1U cabinet space)
Software
- OS: Red Hat Linux 9
- Compiler: GCC (v3.2)
- Parallel Software (MPI libs): MPICH (v1.2.5), LAM (v7.0.3)
- Management: Ganglia/NodeWatch
External Interface
- 1U rackmounted LCD/KVM connected to Master
- Outside LAN connection to Master
- Remote logins to the Master node (SSH)
- Web cluster management (HTTPS)
Cluster Vendor/Integrator
Use
- Excitable media simulations (main use, Dr. Roman Zaritski)
- Modeling of disease dynamics (Dr. Lora Billings)
- Other research projects
- Student research and teaching
Contact: Dr. R. Zaritski
