Case Studies
UCSF HPC Cluster
Satellite Imagery
UCSF HPC Cluster
Near 50% saving over BeeGFS solution
Cluster of 4 nodes, 1.4 PB, $100K
80% capacity utilization
Media Production
Satellite Imagery
UCSF HPC Cluster
Out-perform EMC Isilon S200 Series
60% more capacity
50% higher throughput
Satellite Imagery
Satellite Imagery
Satellite Imagery
Same number of drives, 144
Dual 10 GbE vs. Infiniband
No buddy mirroring
University of California
UCSF Wynton HPC Center
- 1.2 PB storage, estimated cost $192K USD
- Hardware: 4 nodes of 60-bay servers, 2 nodes of metadata servers
- Software: ZFS and BeeGFS
Competitive LeoFS Cluster
- 1.4 PB storage, only $100K USD
- Throughput: read 6.8GB/s, write 10GB/s (asynchronous)
- Hardware: 4 nodes of 36-bay servers
2 Metadata + Storage nodes
- CPU:Intel Xeon E5-2630 V4 x 2
- Motherboard: Supermicro X10DRL-i
- HBA: LSI SAS 9300-8I
- System Disk Drive: 480 GB SSD x 2 + 240 GB SSD x 2
- Storage Disk Drive: 10 TB HDD x 34
- RAM: 128 GB
- Network Port: 4 x 10 GbE
2 Storage only nodes
- CPU:Intel Xeon E5-2630 V4 x 2
- Motherboard: Supermicro X10DRL-i
- HBA: LSI SAS 9300-8I
- System Disk Drive: 240 GB SSD x 2
- Storage Disk Drive: 10 TB HDD x 36
- RAM: 64 GB
- Network Port: 4 x 10 GbE
Animation 3D
EMC Isilon Case Study
- Ten nodes of S200 series
- Raw capacity of 600 TB
- I/O throughput 8 GB/s
Competitive LeoFS Cluster
- Ten nodes of 4U 24-bay storage servers
- Raw capacity of 960 TB
- I/O throughput over 12 GB/s
Geneva Observatory
BeeGFS Case Study
- 4 storage nodes and 2 metadata servers
- 144 drives
- Infiniband
- Effective 800 TB
- I/O 5-8 GB/s
Competitive LeoFS Cluster
- 4 nodes of 4U 36-bay storage servers, 2 metadata servers
- 144 drives
- Dual 10 GbE
- Usable 1PB
- I/O 8-11 GB/s
- No buddy mirroring, 80% capacity utilization
- No single point of failure from drives, nodes or network
- File-level RAID, faster data recovery
- Better ROI