Infrastructure for High Performance Computing (HPC)
High Performance Computing (HPC) infrastructure consists of compute resources (processors, accelerators such as GPUs, RAM, storage and high-speed interconnect) that enable researchers to solve computationally demanding problems. The HR-ZOO infrastructure for HPC consists of:
- compute nodes:
- processor nodes;
- GPU nodes;
- fat nodes with large RAM capacity;
- shared storage;
- login nodes;
- high-speed interconnect;
- software.
Infrastructure for High Throughput Computing (HTC)
High Throughput Computing (HTC) infrastructure provides a cloud computing platform enabling users to deploy virtual machines and applications that require significant compute and storage resources. HTC Cloud provides flexibility in using and running various operating systems and applications, scaling machines on-demand and creating custom network topologies. HTC is based on open systems:
- OpenStack – set of software tools for building and managing cloud computing platforms
- Ceph – distributed object store and file system.
The HR-ZOO infrastructure for HTC is deployed in four data centres. HR-ZOO ZG2 contains compute and storage resources and controller nodes on which the central services are located. Other HR-ZOO data centres contain only computer resources.
Infrastructure for High-Scalability Computing (HSC)
HSC infrastructure provides network, computing and storage resources for cloud computing.
The backbone of the HR-ZOO infrastructure for HSC is the stretched cluster between the HR-ZOO ZG1 and HR-ZOO ZG2 data centres, two
individual high-availability clusters (HA) with mutually synchronously replicated data and the possibility of automated HA failover of the virtualized resources from one location to another.
Disaster recovery centres (DR) are other HR-ZOO data centres:
HR-ZOO OS, HR-ZOO RI and HR-ZOO ST.
Storage Area Network (SAN) infrastructure
The HR-ZOO infrastructure for Storage Area Network (SAN) consists of 3 components that provide users with several options to store and retrieve data, as well as perform backup and recovery:
- HSC storage space is a high-performance system with purpose to store the virtual machines drives and mainly structured data of applications that run on them.
- a unified file and object storage is a system with main purpose to provide users with the ability to store large amounts of data via file and object protocols
- backup of multiple copies of originals of production data in active use that allows data recovery and recovery of active business processes in full operational condition
- archive data by making multiple copies of selected data that are no longer in active use and migrating data from the source storage to the archive storage, which allows retrieving archived data from a certain point in time to find historical information, evidence and knowledge of activities.
Local Area Network (DC LAN) Infrastructure
The HR-ZOO infrastructure for DC LAN is a component that connects all computer and storage components in the HR-ZOO data centres. DC LAN is installed in all HR-ZOO data centres and in addition to HPC, HTC, HSC and SAN, it also connects hosted ICT equipment (EXT) and the CARNET network.
DC LAN consists of: switching layer, routing layer and security layer.
Management and monitoring are integrated through all layers and through all data centres. In order to unify HR-ZOO infrastructures throughout data centres, HR-ZOO transport network is used, and in order to be visible on the Internet, the CARNET network is used as an Internet service provider.
Image source: Designed by starline / Freepik
