What is this service?
Advanced Computing provides computing infrastructure for solving computational resource-demanding problems: Supercomputer Supek and the cloud computing platform Vrančić. Cluster Isabella, HTC Cloud (High-Throughput Computing) computing environment, and the VCL (Virtual Computing Lab) are also available for researchers as part of this service.
Who can use this service?
This service can be used by every natural person fluffing requirements stated in the document Rules of service use:
How can you become a user?
You can submit your application request on this web page: https://computing.srce.hr.
Advanced computing applies state-of-the-art computing infrastructure to help researchers solve resource-demanding problems in various scientific fields such as artificial intelligence, bioinformatics, computational chemistry, climate modelling, seismology and many others.
Resources
Supercomputer Supek provides researchers High Performance Computing (HPC) infrastructure. HPC enables the execution of resource-demanding applications via a large number of CPU cores, graphical processors, RAM, storage and high-speed network. Supek is based on HPE Cray technology and consists of 8384 processor cores, 81 GPUs, and 32 TB of working memory providing a power of 1.25 PFLOPS, which makes it the first petascale supercomputer in Croatia. It is located in the area of the newly built data centre at University Campus Borongaj in Zagreb. Most of the servers are housed in just one HPE Cray EX2500 cabinet, which is a fully integrated supercomputer. The advantage of this system is that 100% of the heat is removed by Direct Liquid Cooling (DLC) which makes it extremely energy efficient. Storage system is HPE ClusterStor 1000 with 580 TB usable capacity and measured bandwidth of 238 GB/s for writing and 450 GB/s for reading.
| CPU servers | GPU servers | Servers with large storage capacity | |
| Total nodes | 52 | 20 | 2 |
| Total CPUs | 6656 | 1280 | 256 |
| Total GPUs | 80 | ||
| Processor | 2 x AMD Epyc 7763 2,45 GHz | 1 x AMD Epyc 7763 2,45 GHz | 2 x AMD Epyc 7763 2,45 GHz |
| CPU | 128 | 64 | 128 |
| Working memory | 256 GB | 512 GB | 4 TB |
| GPU | 4 x NVIDIA A100 with 40 GB memory |
Cloud computing platform Vrančić is intended for a wider set of user applications and systems which require a specific programming environment and significant processor and memory resources. It is useful for researchers who need an adaptable, elastic and versatile computing resource capable of meeting the diverse requirements in different scientific fields. Vrančić is based on OpenStack cloud computing infrastructure providing 11520 processor cores, 16 GPUs and 57 TB of working memory.
On Vrančić, a computing cluster named Padobran has been established, utilizing a portion of Vrančić's resources, specifically 50 CPU nodes, and it also has its own access node with 16 CPU cores. Unlike Supek, it does not have a Slingshot interconnection between nodes and lacks GPU nodes, but it has more RAM and allows job execution for up to 30 days, whereas on Supek, it is only 7 days. Its storage has a total capacity of 464 TB.
| CPU servers | GPU servers | Servers with large storage capacity | |
| Total nodes | 86 | 4 | 2 |
| Total CPUs | 11008 | 256 | 256 |
| Total GPUs | 16 | ||
| Processor | 2 x AMD Epyc 7713 2,0 GHz | 1 x AMD Epyc 7713 2,0 GHz | 2 x AMD Epyc 7713 2,0 GHz |
| CPU | 128 | 64 | 128 |
| Working memory | 512 GB | 512 GB | 2 TB |
| GPU | 4 x NVIDIA A100 with 40 GB memory |
Jupyter platform enables interactive computing with support for multiple programming languages and tools (Python, Julia, Octave, Pluto Notebook, R, RStudio, SageMath, Wolfram Engine, Bash, Markdown). Through Jupyter, an environment with a graphics card is available, and by using Multi Instance GPU technology (MIG), a single A100 GPU card can be used by multiple users for compute intensive applications and programs. For data storage, disk space is available where user data remains saved and is not deleted after rebooting, while for larger amounts of data, the Srce Data Storage and Management System PUH can be used. Depending on the choice of environment and system load, 2 - 12 CPU cores, 6 - 24 GB of RAM, MIG GPU A100, and 10 GB disk are available. From the Jupyter platform, it is possible to easily access other Srce resources that are part of the Advanced Computing service (Supek, Padobran, virtual server Vrančić, etc.). The platform is established on the Vrančić resource.
Galaxy platform s an open and web-based platform for scientists that enables easy parameter setting, running individual tools, as well as using, sharing, analyzing, and visualizing data. The Galaxy platform is based on the principles of FAIR (Findable, Accessible, Interoperable, Reusable) and uses the resources of the Padobran cluster. Using a range of programs and tools available within the Galaxy platform, including BLAST, bwa, fastqc, and many others, researchers can perform analyses such as genome assembly, sequence comparison, gene identification and their functions, and the study of genetic variations.
Support
Support includes consulting for choosing the best-fit resource based on users' needs, assistance with setting up the environment, updates for existing applications, installation and optimization of new scientific applications, as well as assistance in running applications.
Education through customized self-paced and hands-on workshops is also part of this service. The workshops are intended for new users who want to get to know the cluster environment and the basics of the job management system. In addition to general tasks such as job submitting, running specific scientific applications is also demonstrated.
History
Advanced computing in SRCE started with the first cluster Dgrid with 8 single-CPU nodes and cluster distribution Rocks. It was established in 2002. for the needs of integrating Croatia into European project DataGrid led by CERN with the main goal of building a grid for large-scale storage and analysis of data collected by LHC (Large Hadron Collider) experiment.
Dgrid evolves into HPC cluster Isabella on May 8th 2002. – a shared resource of all scientists in Croatia. Since then, Isabella went through over 10 upgrades until it reached its current structure in 2018.
