May we introduce: AISA Infrastructure
After a long time seeing it coming, the AISA Infrastructure is finally maturing.
First, we installed hyperflex nodes alongside an existing installation of the computer science department. These are perfectly suited for setting up virtual machines in an industry grade server environment. It features 3 nodes, doubly redundant SSDs and 120 logical CPU cores, a total of 2.25 TB memory and a storage of 30 TB.
Second, the AISA GPU-Cluster has eventually been installed. It features a total of four nodes:
- 2 nodes, each equipped with Dual-CPU (AMD Epyc 7713 chipset, 64 cores), 1 TB of RAM, roughly 20 TB of high quality SSD storage, and each featuring 8x Nvidia A100 (80 GB RAM each)
- 3 nodes, each equipped with Dual-CPU (AMD Epyc 7513 chipset, 32 cores), 512 GB of RAM, roughly 20 TB of high quality SSD storage, and each featuring 8x Nvidia A40 (48 GB RAM each)
- a dial-in node, equipped with AMD Epyc 7302P CPU (16 cores), 128 GB of RAM, roughly 4 TB of SSD storage
- high speed and low latency communication is guaranteed through Mellanox hardware (InfiniBand 200 Gb/s)
Until access can be rolled out we still need to spend some time on the configuration. We are thrilled to see the first AI+SE+X centered projects put the newest additions to AISA Infrastructure to good use.