PERFORMETRICA HPC CLUSTER QUICKSTART GUIDE
Scope of This Guide
This guide explains how to access and run jobs on the Performetrica HPC Cluster.
It does NOT cover:
-
Basic Linux usage
-
Programming models such as MPI, OpenMP or CUDA
Users are expected to have:
-
Basic Linux command-line knowledge
-
Familiarity with their applications
Getting an HPC Account
Access to Performetrica HPC is managed via the user portal:
👉 https://portal.superbilgisayar.tr/
Steps
- Register via the portal (Online Registration)
- Verify your email and complete activation
- Wait for approval (if required)
- Your HPC account will be automatically provisioned
Accounts may be:
-
Individual (R&D users)
-
Organization-based (universities, companies)
Important Usage Rules (Read Carefully)
Improper usage can negatively impact all users on the system.
Failure to follow these rules may result in account suspension.
Do NOT
- Run computational jobs on login nodes
- Submit large-scale jobs without testing
- Abuse shared resources
Always
- Start with small test jobs
- Use the scheduler (Slurm)
- Monitor your jobs
HPC System Architecture
The Performetrica HPC cluster consists of:
- Login Nodes → login, job submission, light tasks
- Compute Nodes → actual computation
- Storage Systems → shared high-performance storage
Login nodes are NOT for computation.
Connecting to the System
Windows
Use an SSH client such as:
-
MobaXterm
-
PuTTY
After installing the client, connect using: username@login.superbilgisayar.tr
Linux and macOS
Use SSH from terminal:
ssh username@login.superbilgisayar.tr
File System Basics
When you login, you start in your home directory (~).
Basic Commands
pwd # current directory
ls # list files
cd dir # change directory
cd .. # parent directory
File Operations
mv file1 file2 # rename/move
cp file1 dir/ # copy
Data Transfer
Upload (local → cluster)
scp file username@login.superbilgisayar.tr:/path/
Download (cluster → local)
scp username@login.superbilgisayar.tr:/path/file
Editing Files
Use terminal-based editors:
- vi
- nano
- emacs
Software Environment
Available software is managed via modules:
module avail
module load <software>
For additional software:
Submit a request via the portal.
Job Scheduling (Slurm)
All computations MUST be submitted via Slurm.
Example Job Submission
sbatch job.sh
Useful Commands
squeue # list jobs
scancel JOBID # cancel job
sinfo # cluster status
Jobs are scheduled based on:
-
Resource availability
-
Fair usage policies
-
Priority
Storage Policy (Critical)
- HPC is NOT for long-term storage
- Always backup your data externally
- Remove unnecessary files after completion
Old or unused data may be deleted by administrators.
Best Practices
- Test jobs before scaling
- Use appropriate resources (CPU, memory)
- Avoid unnecessary large I/O operations
- Keep your directories organized
Support
For help and support: