Design and management of large-scale computing infrastructures
The Intelligent Systems Group (ISG, http://www.sc.ehu.es/isg) of the School of Informatics of the University of the Basque Country UPV/EHU (http://ehu.eus/if) has an open position as part-time junior researcher. It will be a one year contract (starting January 1st 2018), renewable, with a net salary of 890€ (approximately) and a 50% dedication. This position tries to attract recent graduates who want to pursue a PhD in Computer Science, in the areas of work described below.
The candidate should have a degree in Computer Science, with specialization in high-performance computing and networking and experience in performance analysis (via mathematical modelling and simulation) and a working knowledge of machine learning methods. Good programming skills in languages such as C and Python are also required. A good command of the English language is essential.
The work will focus on the design and management of large-scale computing infrastructures, such as cloud datacenters. In particular, we want to optimize the way different applications share the infrastructure, in order to maximize application performance while minimizing infrastructure cost, including energy. As the size of the datacenter grows, failures are not rare events, and applications must be resilient to these failures. Thus, cloud schedulers and applications must be designed with mechanisms to provide fault tolerance.
To that extent, we need to characterize cloud applications in these aspects:
- The way they arrive to – leave the datacenter, and the interaction with the scheduling mechanisms managing it.
- The way they request resources and release them, including auto-scaling.
- The way they communicate internally and with the external world, and the relationships between the use of the network and the performance of the applications.
- The way they may be affected by hardware / software failures and the mechanisms used to avoid or minimize their effects.
We claim that the interconnection network plays a fundamental role in achieving the above-mentioned objectives. Additionally, the infrastructure must provide isolation between applications, and also mechanisms to allow the deployment of sophisticated network functions such as load balancing, firewalling, multipath routing, etc. For this reason we will study the use of SDN (Software Defined Networks) technologies, combined with datacenter-specific topologies and addressing schemes, as a way of implementing datacenter-optimized networks.
The research work will be carried out using a combination of different tools and methodologies, including (but not restricted to) mathematical modeling, machine learning techniques, discrete-event simulation, prototyping with emulators (virtual environments), and small-scale hardware prototypes.
For more information, contact with Prof. Jose Miguel-Alonso (firstname.lastname@example.org) as soon as possible. Applications should be made through the UPV/EHU’s web site before November 14, 2017.