JAKAR Cluster
Overview
The JAKAR cluster, UTEP's central High Performance Computing resource, is comprised of 72 compute nodes, providing a total of 2,880 CPU cores and over 28 TB of RAM. Nodes are categorized into partitions according to research disciplines and hardware investments.
Cluster Breakdown:
- Total Nodes: 72
- Total CPUs: 2,880
- Total RAM: 28 TB
- Partitions: 12
| Specification | Details |
|---|---|
| Manufacturer | Lenovo |
| Model | ThinkSystem SD530 & SR630 |
| Processors | 2× Intel(R) Xeon(R) Gold 6230 |
| Cores/Node | 40 (20 per socket) |
| Clock Rate | 2.40GHz (Base Frequency) |
| Memory/Node | 187Gi to 755Gi |
Student Academic HPC Resource
The JAKAR cluster has allocated resources for academic purposes to help students learn parallel programming and gain experience working with the High-Performance Cluster (HPC). These resources are accessible by submitting jobs through the SLURM scheduler using the Academic allocation. This allocation is primarily reserved for students to submit jobs and receive results almost immediately.
Resource Allocation
The Academic allocation will use the small nodes, with fair-share access to ensure that all students and class work have equal time. This allocation can be requested by professors or students through a service desk call.
Cluster Specifications
The JAKAR cluster is based on the Cascade Lake family with the Intel Xeon Gold 6230. It includes: - Total Nodes: 9 - Cores per Node: 40 - Memory per Node: 192GB
Total Academic Allocation
- Total Cores: 360
- Total Memory: 1,728GB
- Wall-clock Limit: 4 hours (maximum job duration)
Class/Student Allocations
- Specific Submission Queue
- Cores Allocated: 4 (customizable based on class needs)
- Memory per Core: 9.6GB
- Wall-clock Limit: 4 hours
Hint
If you need an account for the Jakar cluster, uses the below links.
Access Information
JAKAR Evolution
The JAKAR cluster has evolved over several generations, each marked by major system upgrades and architectural improvements to support cutting-edge research.
JAKAR 1.0 — Foundation Phase
- OS: CentOS 6
- Storage: Network File System (NFS)
- Scheduler: IBM LSF
- Allocation: Static, single general partition
- User Management: Manual database
- Network: 10Gb Ethernet
- Hardware: Intel Xeon (12 cores per dual socket), 64 GB RAM
JAKAR 1.5 — Expansion and Refinement
- OS: CentOS 7
- Storage: GPFS
- Scheduler: SLURM
- Allocation: Static partitions by investment
- User Management: Script-based
- Hardware: Intel Xeon (20 cores per dual socket), 384 GB and 756 GB RAM
JAKAR 2.0 — Scaling Up
- OS: CentOS 7
- Storage: GPFS
- Scheduler: SLURM v2
- Software Stack: OpenHPC v2, COS 7.x
- Network: 10Gb with planned upgrade paths
- Hardware: Continued use of 20-core Xeons with 384 GB / 756 GB RAM
JAKAR 3.0 — Current Generation
- OS: RHEL 9.2
- Storage: Upgraded GPFS for RHEL 9
- Scheduler: SLURM v3
- Software Stack: OpenHPC v3
- Allocation: Research on dynamic allocation, incorporating quotas and investment
- User Management: Local account management
- Hardware: Intel Xeon (20 cores per dual socket), 384 GB / 784 GB RAM
- Network: 10Gb; transition to InfiniBand under consideration
- Note: Hardware expansion of SD530 nodes ended in September 2023