Germany - server

Description

The evaluation of very large amounts of data, as generated by modern analysis devices, requires such a high computing power, which can only be achieved by parallelizing the algorithms and heuristics on several levels. This requires a hardware platform consisting of multiple multithreaded CPU servers, each with multiple GPUs, which can exchange over a high-speed network. The data analysis cluster is intended to overcome previous limitations. The services of the contractor are:- Delivery of the equipment including required accessories such as network optics (also for the existing Cisco Nexus 9300), cables, licenses and small material (e.g. Cable ties, labels, Velcro, tape),- free delivery to point of use,- hardware and operating system installation,- necessary cabling,- integration into the existing infrastructure of the University of Leipzig, such as SLURM queue system and storage and user management, Puppet and Foreman configuration management system,- setup of an Infiniband EDR network,- installation of the BeeGFS storage system,- setup of SLURM job workflows including all services, which start virtual machines with Windows 10 with the aid of Libvirt, QEMU and KVM, transfer GPU resources to the VM using SR-IOV and release them again upon completion,- testing, functional testing, benchmarking of the entire system,- automation of cluster administration/installation with the aid of Foreman and Puppet,- on-site training/instruction in operation and administration of the cluster system,- free German-language hotline,- troubleshooting by telephone or e-mail.

Opportunity closing date

27 May 2019

Value of contract

£0-100k