Magus02 — HPC Knowledge Base

Latest revision as of 06:25, 20 April 2026

Shiv Nadar Institution of Eminence — HPC Infrastructure

Magus02 HPC
Knowledge Base

Your complete reference for accessing, using, and getting the most out of the Magus02 High Performance Computing cluster at SNU. Whether you are a first-time user or an experienced HPC programmer, this wiki has you covered.

HOST: magus02.snu.edu.in ● SYSTEM OPERATIONAL SCHEDULER: Slurm

🚀 Quick Navigation

⚡ New to Magus02? Start here → Getting Started Guide Everything from account access and SSH login to running your first job — in one guide.
📘 Basics Getting Started · First steps on Magus02 Linux Shell · Commands and filesystem navigation SSH Access · Connecting securely to the cluster File Transfer · scp, rsync, WinSCP, FileZilla Environment Modules · Loading software on HPC Scheduling Basics · How the batch system works	⚙️ HPC User Guide Compilers · GCC, Intel, LLVM, PGI Batch Scheduler · Submitting and managing jobs OpenMP Usage · Shared-memory parallelism MPI Usage · Distributed-memory parallelism NUMA Architecture · Memory locality and pinning FAQ: Batch Jobs · Common problems and solutions

🖥️ About Magus02

Magus02 is the primary HPC cluster at Shiv Nadar Institution of Eminence, Delhi NCR, hosted within the Department of Chemical Engineering. It provides researchers, faculty, and students with large-scale parallel computing resources for a wide range of workloads including computational chemistry, fluid dynamics, machine learning, and data analysis.

Cluster at a Glance

TOTAL CORES	PEAK PERFORMANCE	COMPUTE NODES	GPU NODES	TOTAL RAM	STORAGE
9,064	238 TFlops	57	2 (Tesla A100×4)	~42 TB	750 TB

Node Type	Processor	RAM / Node	Cores / Node	GPU	Count	Total RAM	Total Cores
Compute Node	`AMD EPYC 7763`	512 GB	128	—	57	29,184 GB	7,296
High Memory Node	`AMD EPYC 7543`	1,024 GB	64	—	12	12,288 GB	768
GPU Node	`Intel Xeon Platinum 8368`	512 GB	64	Tesla A100 ×4	2	1,024 GB	128
Storage	`750 TB` Parallel Filesystem

Queue Summary

Queue	Node Type	Max Cores/Job	Max Nodes	Wall Time	Use Case
medium	Compute	128	77	1 week	General multi-node jobs
small	Compute	32	16	1 month	Testing and small runs
high_mem	High Memory	256	12	2 weeks	Gaussian, large in-memory jobs
large	Compute	256	24	2 weeks	Large parallel production runs
GPU	GPU	16	4	1 week	CUDA/GPU workloads
low/long	Compute	1024	All	1 month	Long, low-priority jobs

→ See Magus02 Cluster Page for full queue details, connection instructions, and application list.

📚 Documentation Sections

Basics (New Users)

🖥️ Getting Started Access, login, VPN, first steps on the cluster.	⌨️ Linux Shell Essential commands for navigation, files, and utilities.	🔒 SSH Access Connecting from Linux, macOS, and Windows.
📁 File Transfer scp, rsync, FTP/SFTP, WinSCP, FileZilla.	📦 Environment Modules Loading compilers, MPI, and scientific software.	📋 Scheduling Basics Batch queues, job lifecycle, and scheduling algorithms.

HPC User Guide

🔧 Compilers GCC, Intel icc/ifort, LLVM/clang, PGI/NVIDIA.	📤 Batch Scheduler Job types: serial, OpenMP, MPI, hybrid, array, chain.	🔀 OpenMP Usage Compiling and running shared-memory parallel programs.
🌐 MPI Usage Open MPI and Intel MPI: compile, run, bind.	🏗️ NUMA Architecture Non-uniform memory access, thread pinning, first-touch.	❓ FAQ: Batch Jobs Troubleshooting common Slurm job issues.

🔗 Resources & Support

📞 Getting Help

Contact the HPC Administration Team for access requests, technical issues, and software installation.

Email: [email protected]
IT Helpdesk: Submit a ticket via the SNU IT portal

📄 Downloads

⬇ Slurm / Linux Command Reference (PDF)

All major Slurm commands, job submission flags, and useful Linux utilities in a single printable reference sheet.

⚠️ External Access Reminder: Accessing Magus02 from outside the SNU campus network requires a VPN connection. Contact the HPC admin team or scan the QR barcode on your welcome email to request VPN access. See SSH — External VPN Access for details.

Maintained by HPC Administration · Shiv Nadar Institution of Eminence, Delhi NCR · wiki.snu.edu.in