An HPC cluster must not be seen as just high-end hardware, otherwise you run the risk of it being just an empty and useless shell. Software tools are therefore necessary to facilitate both its use and its management and monitoring.
The complexity of these systems can pose a challenge for administrators and users. BECHTLE is there to help you make the right technological choices that match the goals you want to achieve.
BECHTLE can offer you a wide range of solutions to suit your needs, but also your IT environment. These solutions can be divided into two categories:
In both cases, we are happy to support you in all phases of implementing and using your HPC cluster.
Below you will find the most common requirements in the area of HPC as well as examples of solutions that we can offer you.
Phone: +33 3 67 07 97 38
Phone: +33 3 67 07 97 38
BeeGFS is a parallel file system specially developed and optimised for high performance computing. A parallel file system splits the data into several parts and distributes them over the storage space. The aim is to significantly improve simultaneous access and performance compared to a "classic" file system.
BeeGFS can be downloaded and installed without user fees. This product is used in many top 500 supercomputers. BeeGFS runs on every Linux computer. The product is composed of various services: Storageserver, Metadataserver, Client…
Professional support is possible This is provided by ThinkParQ.
The main advantages of BeeGFS:
Scality is a French vendor that offers a high-performance software storage solution to meet the demand for a scalable and cost-effective multi-user storage solution.
The solution consolidates all unstructured data into a shared memory for addressing: Backup (the solution is certified for all backup solutions including VEEAM) and long-term archiving; file synchronisation and sharing; Big Data, AI and Analytics; video surveillance…
IT teams can easily take care of multiple departments and services with the agility of public cloud solutions - all in your own data centres. At the same time, they have the option of managing your data (life cycle, research, cost optimisation, etc.) in large public clouds (Azure, Google, AWS, etc.) if required.
The main advantages of the solution:
Bright Cluster Manager, developed by Bright Computing, enables the provision and management of HPC clusters. This tool provides a uniform interface for hardware, operating system, HPC software and users.
Administrators can quickly install clusters and keep them in optimal operating condition throughout their operational cycle.
The main benefits of Bright Cluster Manager:
OpenHPC is an open source toolset for HPC clusters. These tools enable cluster-based installation, administration, resource management, software development, libraries, and metrics monitoring and analysis. The modular structure of OpenHPC's architecture gives users the opportunity to make a selection from the available components.
Some examples of software available in OpenHPC repositories: xCAT, BeeGFS, MPICH, PBS Pro, …
OpenHPC is installed on a master server on which the operating system is already installed.
NICE DCV is a technology developed by NICE SOFTWARE that enables remote access to 2D/3D applications and desktops.
The DCV protocol is able to adapt to heterogeneous networks (LAN, WAN, VPN) to optimise the user experience according to latency and bandwidth.
The principle of this technology is simple: On a server equipped with one or more graphics cards, a 2D/3D application sends data that is converted into images by the graphics cards. The images are then compressed and sent to clients over the network. These clients then only need a simple, lean application or even just an internet browser to receive and display the graphic data. The rendering and user experience is very similar to local use.
NICE DCV can connect to Windows or Linux desktops for OpenGL and DirectX applications. It is also possible to split a GPU into multiple partitions and assign each partition to a different user (NVIDIA cards required).
SLURM is an open source solution for managing the resources of an HPC cluster. The fail-safe solution supports clusters of different sizes - from 2 to several thousand nodes.
The 3 key functions are:
PBS Professional optimises job scheduling and workload in an HPC environment.
The main features of PBS Professional are:
Nagios is known as the best server monitoring software on the market. This style of monitoring is characterised by the fact that a client agent is not needed. More than 5000 plug-ins are available to customise the environment to personal needs.
Nagios is available in two editions:
The main features of Nagios are:
Ganglia is a distributed monitoring tool capable of displaying real-time metrics (CPU utilisation, RAM, network, ) from the various components of an HPC cluster.…
It is available directly from the packages of the most commonly used Linux distributions.
Ganglia also runs on very large clusters, but requires little space. The tool can be installed on a wide range of architectures and is used on more than 500 clusters around the world. It also offers the possibility of connecting clusters in different geographical locations.
The Ganglia ecosystem includes 2 services for sending and collecting data, a PHP web interface and some utilities.