Internship: Virtual development environment with GPU and Infiniband support

To support the development of parallel algorithms, a heterogeneous cluster is operated by IPE. Eight workstations in the student cluster are equipped with GPUs and Infiniband adapters and are connected to the IPE storage servers. On one hand these nodes are used to test GPU-accelerated parallel applications and in parallel the students are using the workstations as their desktop computers. To get reliable results, we need to keep the same software on all cluster nodes. To allow users flexibility in the selection of operating systems and applications, virtualization technology should be used. It is goal of this student work to evaluate existing technologies and propose a solution that ensures uninterrupted cluster operation and allows flexible choice of the users desktop configuration.

The main requirements are
• Transparency: the workstations should start the selected virtual machine at boot time automatically and the virtualization should be completely hidden from the users.
• GPU-passthrough: the GPUs need to be fully accessible by the virtual machines.
• VM Migration: Ii should be possible to run virtual machines on any of the cluster nodes interchangeably.
• Recovery: Disk snapshots should be used to allow easy recovery in case of failures.
• Resource isolation: cluster applications should be IPE Computing Infrastructure able to use exclusively all required resources. It is necessary to ensure that desktop users can't influence Performance measurements.
• Boostraping: a bootstraping procedure for creating new virtual machines tailored for specific projects should be proposed.


Weitere Informationen

Unternehmen
Helmholtz Gemeinschaft
Standort
Karlsruhe, Baden-Württemberg, Deutschland
Anstellungsart
Praktikum
Vertiegungsgebiet
Forschung und Entwicklung
Ansprechpartner/in
Fachliche Auskünfte erteilt Ihnen gerne Suren Chilingaryan, IPE, +49 721/608 26579 (Mail: suren.chilingaryan@kit.edu) and Timo Dritschler, IPE, +49 721/608 25693, (Mail: timo.dritschler@kit.edu)

Bewerbung:
Bitte senden Sie Ihre Bewerbung online unter Angabe der Stellenausschreibungsnummer IPE 01-16 an Frau Berger, Berufliche Ausbildung, Telefon 0721 608-25184.
Bei entsprechender Eignung werden schwerbehinderte Bewerber/innen bevorzugt berücksichtigt.
Anforderungen
Required Skills:
Very good knowledge of Linux and the scripting languages. Experience in virtualization technologies is a plus. The student should be ready to detect and solve hardware and software problems.

Experience gained:
Cluster management, virtualization, GPU computing.
Link zum Jobangebot
zum Jobangebot



TIPP: Dein Profil wird dem Unternehmen übermittelt. Erziele einen besseren Eindruck, indem Du es vollständig ausfüllst.