Kazakhstan National Research & Data Platform
KazNRDP
One path from a scientific question to data, compute, and a result. Farabi Supercomputer Center is the first core.
DATA · HPC · AI · RESEARCH SERVICES
From a question to a result
KazNRDP shortens that path. The researcher does not hunt for files, then beg for machines, then move terabytes by hand.
- 01
Question
Climate, genomics, geology, materials, or a model to train. The work starts here.
- 02
Data
One catalog says what exists, who owns it, and how to use it. The files can stay where they are.
- 03
Compute
CPU, GPU, a notebook, or a container runs beside that data. Move the job, or move the data — whichever is smaller.
- 04
Result
A forecast, a map, a trained model, or a figure. Not another copy of the raw archive.
A core, not a warehouse
Machines and data can stay at other universities and institutes. KazNRDP joins them with shared access, catalogs, and identity.
- Peak Performance
- ~2 PFLOPS
- CPU Compute Nodes
- 400+ Nodes
- NVIDIA A10 Accelerators
- 30 GPU Nodes
- SSD next to the machines
- 384 TB
- Archive, after the job
- 6 PB
The SSD holds the job. The archive holds what the job already produced.
Active work stays on 384 TB of fast disk beside the machines. Finished collections move to the 6 PB archive.
384 TB
SSD next to the machines
- Jobs that are running
- Datasets loaded for training
- Virtual machines
- Scratch files from the current run
6 PB
Archive, after the job
- Satellite scenes
- Climate series
- Genomes
- Finished project archives
How it grows
- Stage 1
Core
Portal, identity, catalog, and Farabi CPU and GPU.
- Stage 2
Data
Outside datasets and the first science gateways.
- Stage 3
Compute
Other organizations join the compute pool.
- Stage 4
Nation
Research, data, HPC, and AI as one system.