Compute Infrastructure
HPC Cluster Architecture¶
TU HPC is structured into multiple compute nodes, ensuring scalability and efficiency. The key components of the infrastructure include:
1. Compute Nodes¶
TU HPC consists of multiple compute nodes, categorized into different partitions:
- Normal Partition: General-purpose nodes for CPU-based computing.
- Protein Partition: Dedicated GPU node for high-performance computing.
- FPLO Partition: Nodes configured for specific research applications.
Node Configuration¶
- Total Compute Nodes: 10
- CPU Architecture: x86_64
- Processor Model: Intel(R) Xeon(R) CPU E5-2680 v2 @ 2.80GHz
- Total CPU Cores: 20 per node (10 cores per socket, 2 sockets per node)
- Threading: 1 thread per core
- NUMA Nodes: 2 (Ensuring optimized memory access)
2. Job Scheduling System¶
- SLURM-based workload manager for efficient job scheduling and resource allocation.
- Supports batch and interactive job submission.
- Users can specify CPU, GPU, memory, and time requirements for optimal usage.
3. GPU Infrastructure¶
- Dedicated GPU Node: Available under the Protein Partition.
- Supports CUDA and OpenCL-based parallel computing.
- Optimized for deep learning, molecular dynamics, and AI research.
4. Storage & File Systems¶
TU HPC provides a multi-tiered storage system to support computational research:
- Root Storage:
/dev/mapper/centos-root
– 50GB (OS & system files) - Software Storage:
/dev/sdb1
– 470GB (Installed software and modules) - Scratch Storage:
/mnt/storage0
– 1.9TB (High-speed temporary storage) - Home Directory:
/home
– 1.8TB (User personal storage) - Memory File System:
tmpfs
– 126GB (High-speed temporary memory storage)
5. Networking & Connectivity¶
- High-speed interconnect for low-latency data transfer between nodes.
- Secure SSH access for remote login.
Performance & Scalability¶
The HPC infrastructure at TU is designed to support high-throughput computing (HTC) and massively parallel processing (MPP). It provides:
- Scalability to accommodate growing computational demands.
- Resource optimization to maximize job efficiency.
- Flexible configurations for different research requirements.
Supported Workloads¶
The compute infrastructure supports a wide range of research activities, including:
- Computational Physics & Chemistry
- Machine Learning & Deep Learning
- Molecular Dynamics Simulations
- Weather & Climate Modeling
- Structural Engineering & Finite Element Analysis
- High-Performance Data Analytics
Future Upgrades¶
TU HPC is continuously expanding with:
- More GPU nodes for AI and scientific computing.
- Increased storage capacity for big data applications.
- Enhanced networking and security to ensure seamless operations.
For more details on how to access and utilize TU HPC resources, visit the Getting Started Guide.
Empowering research with high-performance computing! 🚀