Hardware
We provide multiple systems for working with sensitive data:
-
Verdi: General purpose system with large long term storage coupled to compute resources (both CPU and GPU nodes).
-
DGX-2: Expert system for training AI-models based on NVIDIA’s DGX-2 system
Verdi
Verdi is the current incarnation of our Data Science Platform, supporting use of both sensitive and non-sensitive data aimed at users of a wide range of expertise. It’s intended to enable long term data retention - allowing data to be accessed and worked with during periods of lower activity and making it easy to do compute at high end e-infrastructure provided by NAISS.
It consists of two main parts: CPU compute and GPU compute. The cluster has a CEPH-based storage with total capacity of 3.2 PB on Hard Drives and 153 TB Solid State Drives.
The nodes of the cluster is connected by dual 25 Gigabit Ethernet for all nodes.
Field | CPU Compute | GPU Compute L4 | GPU Compute L40s |
---|---|---|---|
Number of nodes | 40 | 6 | 2 |
Storage Devices per node | 6.4 TB NVMe SSD | 6.4 TB NVMe SSD | 6.4 TB NVMe SSD |
Memory (RAM) per node | 1024 GB | 512 GB | 512 GB |
CPUs per node | 32 Cores (2 Intel® Xeon® Gold 6526Y 2.8G) | 32 CPU cores (2 Intel Xeon Gold 5416S 2G) | 32 CPU cores (2 Intel Xeon Gold 5416S 2G) |
GPUs per node | N/A | 4 NVIDIA L4 | 4 NVIDIA L40s |
DGX-2
The DGX-2 system is based on NVIDIA’s system of the same name. The system consists of a DGX-2 server with 15 V100 (32 GB) GPUs connected to a storage server with 10 Gigabit Ethernet. The machine used in virtual machines which divides the resources in quarters (3 or 4 GPUs depending on block). The full specifications of the machines are:
- 15 NVIDIA Tesla V100 (32 GB) GPUs
- Dual Intel Xeon Platinum 8168, 2.7 GHz, 24-cores
- 1.5TB system memory (RAM)
- Storage: 30TB (8X 3.84TB) NVME SSDs
Contact
Our system design team:
![]() |
Pontus Freyhult IT Architect System design support, Policy support, Management, System development |
![]() |
Joel Hedlund Head of AIDA Data Hub Management, Policy support, System design support, Data sharing, Economy |
Please contact us for support or more information!