We formalize the problem of scheduling jobs on a set of machines, each having a fixed number of resources, and a probability to be alive, computed according to some probability distribution, e.g., random walk. The goal is to maximize the utilization. Several heuristics are designed to efficiently schedule jobs, cleverly deciding which machines to use (trade-off between the load and the probability that the machine will not survive until job completion). We will present simulation results, based on real traces, to compare the performance of various heuristics.

The management and allocation of resources to users in HPC infrastructures often relies on the RJMS. One key component for an optimized resource allocation, with respect to some objectives, is the scheduler. Scheduling theory is interesting as it provides algorithms with performance guarantees. These guarantees come at the cost of tedious and complex modeling effort. The growing complexity of nowadays and future platforms (hardware heterogeneity, memory/bandwidth/energy constraints) do push to its limit the scheduling approach. Taking into account this new challenges either requires the design of new overly complex models, or exhibits the crudeness of the existing models. In a sense, the scheduling approach fails to capture the dynamic aspects of the platforms. From the control theory point of view, scheduling algorithms are open-loop systems: the actual state of the platform is not fed back into the decision process. By closing the loop and using some control theory results/techniques, we propose to study how to combine both techniques. This study would take place at various levels: from theory to actual applications.

Energy usage is now a major concern in high performance computing (HPC) due to the ecological impact of electricity production, but also to its cost. To optimize the energy efficiency of supercomputers, researchers usually base their work on models, as accessing actual platform is complex and costly. One of the most studied method consists in changing the processor frequency, but it impacts power, performance and energy in a complex way. We propose to bridge the gap between the theoretical and the practical approaches. To do so we propose a multi cluster, multi application model accurately describing from a theoretical point of view the power and performance of applications subject to DVFS (dynamic voltage and frequency scaling). We also show how to use it on a runtime system with a minimal overhead, using only a few hardware performance counters and RAPL. We also investigate the complexity of reproducible experiments and how to realistically simulate datacenters using the Replay with Feedback methodology.

This talk is about a problem of scheduling tasks with precedence constraints on an Edge/Cloud platform partially powered with renewable energy. I will present the model of the platform and tasks considered, and how we designed different greedy algorithms and other algorithms using Constraint Programming to make the allocation decisions.

As the gap between compute and I/O performance tends to grow, modern High-Performance Computing (HPC) architectures include a new resource type: an intermediate persistent fast memory layer, called burst buffers. This is just one of many kinds of renewable resources which are orthogonal to the processors themselves, such as network bandwidth or software licenses. Ignoring orthogonal resources while making scheduling decisions just for processors may lead to unplanned delays of jobs of which resource requirements cannot be immediately satisfied. We focus on a classic problem of makespan minimization for parallel-machine scheduling of independent sequential jobs with additional requirements on the amount of a single renewable orthogonal resource. We present an easily-implementable log-linear algorithm that we prove is (2+5/6)-approximation.

Most of the theoretical work to study scheduling algorithms is based on very naive models on how the resources behave. While this is fine for most scheduling problems, this is very detrimental when studying distributed systems (HPC centers, datacenters...) where the applications behave very differently depending on where they are executed, but also depending on how all the shared resources (computers, network components, shared file systems) are used. This talks is about Batsim, an infrastructure simulator made to study resource management techniques and scheduling algorithms on distributed systems. Batsim's main architecture is presented, an example instantiation of platform and job models is given, then ongoing architectural changes on Batsim are presented. Future directions are discussed at the end.

In a smart city, the number of users can be as many as the entire city's population, with applications often requiring low response times. Edge and fog computing platforms emerge as viable solutions to meet these demands due to their geographically distributed nature and proximity to users compared to traditional cloud data centers. Moreover, their lower power requirements (compared to energy-hungry data centers) make integrating renewable energy sources into their operations interesting. In this work, we adopt an opportunistic computing approach to use the existing computational infrastructure in the city. Therefore, significant computational power can be harnessed, while we also avoid manufacturing new servers and reduce the environmental impact of the smart-city. In our context, we consider the smart-city application of estimating road traffic that is essential for mobility applications. Our research problem consists of scheduling the tasks in the edge/fog computational infrastructure to achieve Quality of Service goals, such as minimizing the response time, and maximizing the usage of low-carbon energy, as we also assume that the hosts have renewable infrastructure (solar panels and batteries). We will present preliminary results exploring the strategies we are investigating to improve the quality of service and minimize the operational carbon footprint. This work was carried out within the MaaS action of the VILAGIL project, a project co-financed by Toulouse Métropole and France 2030 as part of the Territoires d'innovation program operated by the Banque des territoires.

Contract scheduling is a widely studied framework for designing real-time systems with interruptible capabilities. Previous work has showed that a prediction on the interruption time can help improve the performance of contract-based systems, however it has relied on a single prediction that is provided by a deterministic oracle. In this work, we introduce and study more general and realistic learning-augmented settings in which the prediction is in the form of a probability distribution, or it is given as a set of multiple possible interruption times. For both prediction settings, we design and analyze schedules which perform optimally if the prediction is accurate, while simultaneously guaranteeing the best worst-case performance if the prediction is adversarial. We also provide evidence that the resulting system is robust to prediction errors in the distributional setting. Last, we present an experimental evaluation that confirms the theoretical findings, and illustrates the performance improvements that can be attained in practice.