4.1.1 Algorithms

The most common batch scheduling policy is to consider the jobs according to the First Come First Served order (FCFS) with backfilling (BF). BF is the most widely used policy due to its easy and robust implementation and known benefits such as high system utilization. It is well-known that this strategy does not optimize any sophisticated function, but it is simple to implement and it guarantees that there is no starvation (i.e. every job will be scheduled at some moment).

More advanced algorithms are seldom used on production platforms due to both the gap between theoretical models and practical systems and speed constraints. When looking at theoretical scheduling problems, the generally accepted goal is to provide polynomial algorithms (in the number of submitted jobs and the number of involved computing units). However, with millions of processing cores where every process and data transfer have to be individually scheduled, polynomial algorithms are prohibitive as soon as the polynomial degree is too large. The model of parallel tasks simplifies this problem by bundling many threads and communications into single boxes, either rigid, rectangular or malleable. Especially malleable tasks capture the dynamicity of the execution. Yet these models are ill-adapted to heterogeneous platforms, as the running time depends on more than simply the number of allotted resources, and some of the common underlying assumptions on the speed-up functions (such as monotony or concavity) are most often only partially verified.

In practice, the job execution times depend on their allocation (due to communication interferences and heterogeneity in both computation and communication), while theoretical models of parallel jobs usually consider jobs as black boxes with a fixed (maximum) execution time. Though interesting and powerful, the classical models (namely, synchronous PRAM model, delay, LogP) and their variants (such as hierarchical delay), are not well-suited to large scale parallelism on platforms where the cost of moving data is significant, non uniform and may change over time. Recent studies are still refining such models in order to take into account communication contentions more accurately while remaining tractable enough to provide a useful tool for algorithm design.

Today, all algorithms in use in production systems are oblivious to communications. One of our main goals is to design a new generation of scheduling algorithms fitting more closely job schedules according to platform topologies.

4.1.2 Locality Aware Allocations

Recently, we developed modifications of the standard back-filling algorithm taking into account platform topologies. The proposed algorithms take into account locality and contiguity in order to hide communication patterns within parallel tasks. The main result here is to establish good lower bounds and small approximation ratios for policies respecting the locality constraints. The algorithms work in an online fashion, improving the global behavior of the system while still keeping a low running time. These improvements rely mainly on our past experience in designing approximation algorithms. Instead of relying on complex networking models and communication patterns for estimating execution times, the communications are disconnected from the execution time. Then, the scheduling problem leads to a trade-off: optimizing locality of communications on one side and a performance objective (like the makespan or stretch) on the other side.

In the perspective of taking care of locality, other ongoing works include the study of schedulers for platforms whose interconnection network is a static structured topology (like the 3D-torus of the BlueWaters platform we work on in collaboration with the Argonne National Laboratory). One main characteristic of this 3D-torus platform is to provide I/O nodes at specific locations in the topology. Applications generate and access specific data and are thus bounded to specific I/O nodes. Resource allocations are constrained in a strong and unusual way. This problem is close for actual hierarchical platforms. The scheduler needs to compute a schedule such that I/O nodes requirements are filled for each application while at the same time avoiding communication interferences. Moreover, extra constraints can arise for applications requiring accelerators that are gathered on the nodes at the edge of the network topology.

While current results are encouraging, they are however limited in performance by the low amount of information available to the scheduler. We look forward to extend ongoing work by progressively increasing application and network knowledge (by technical mechanisms like profiling or monitoring or by more sophisticated methods like learning). It is also important to anticipate on application resource usage in terms of compute units, memory as well as network and I/Os to efficiently schedule a mix of applications with different profiles. For instance, a simple solution is to partition the jobs as "communication intensive" or "low communications". Such a tag could be achieved by the users them selves or obtained by learning techniques. We could then schedule low communications jobs using leftover spaces while taking care of high communication jobs. More sophisticated options are possible, for instance those that use more detailed communication patterns and networking models. Such options would leverage the work proposed in Section 4.2 for gathering application traces.

4.1.3 Data-Centric Processing

Exascale computing is shifting away from the traditional compute-centric models to a more data-centric one. This is driven by the evolving nature of large scale distributed computing, no longer dominated by pure computations but also by the need to handle and analyze large volumes of data. These data can be large databases of results, data streamed from a running application or another scientific instrument (collider for instance). These new workloads call for specific resource allocation strategies.

Data movements and storage are expected to be a major energy and performance bottleneck on next generation platforms. Storage architectures are also evolving, the standard centralized parallel file system being complemented with local persistent storage (Burst Buffers, NVRAM). Thus, one data producer can stage data on some nodes' local storage, requiring to schedule close by the associated analytics tasks to limit data movements. This kind of configuration, often referred as in-situ analytics, is expected to become common as it enables to switch from the traditional I/O intensive workflow (batch-processing followed by post mortem analysis and visualization) to a more storage conscious approach where data are processed as closely as possible to where and when they are produced (in-situ processing is addressed in details in section 4.3). By reducing data movements and scheduling the extra processing on resources not fully exploited yet, in-situ processing is expected to have also a significant positive energetic impact. Analytics codes can be executed in the same nodes than the application, often on dedicated cores commonly called helper cores, or on dedicated nodes called stagging nodes. The results are either forwarded to the users for visualization or saved to disk through I/O nodes. In-situ analytics can also take benefit of node local disks or burst buffers to reduce data movements. Future job scheduling strategies should take into account in-situ processes in addition to the job allocation to optimize both energy consumption and execution time. On the one hand, this problem can be reduced to an allocation problem of extra asynchronous tasks to idle computing units. But on the other hand, embedding analytics in applications brings extra difficulties by making the application more heterogeneous and imposing more constraints (data affinity) on the required resources. Thus, the main point here is to develop efficient algorithms for dealing with heterogeneity without increasing the global computational cost.

4.1.4 Learning

Another important issue is to adapt the job management system to deal with the bad effects of uncertainties, which may be catastrophic in large scale heterogeneous HPC platforms (jobs delayed arbitrarly far or jobs killed). A natural question is then: is it possible to have a good estimation of the job and platform parameters in order to be able to obtain a better scheduling ? Many important parameters (like the number or type of required resources or the estimated running time of the jobs) are asked to the users when they submit their jobs. However, some of these values are not accurate and in many cases, they are not even provided by the end-users. In DataMove, we propose to study new methods for a better prediction of the characteristics of the jobs and their execution in order to improve the optimization process. In particular, the methods well-studied in the field of big data (in supervised Machine Learning, like classical regression methods, Support Vector Methods, random forests, learning to rank techniques or deep learning) could and must be used to improve job scheduling in large scale HPC platforms. This topic received a great attention recently in the field of parallel and distributed processing. A preliminary study has been done recently by our team with the target of predicting the job running times (called wall times). We succeeded to improve significantly in average the reference EASY Back Filling algorithm by estimating the wall time of the jobs, however, this method leads to big delay for the stretch of few jobs. Even if we succeed in determining more precisely hidden parameters, like the wall time of the jobs, this is not enough to determine an optimized solution. The shift is not only to learn on dedicated parameters but also on the scheduling policy. The data collected from the accounting and profiling of jobs can be used to better understand the needs of the jobs and through learning to propose adaptations for future submissions. The goal is to propose extensions to further improve the job scheduling and improve the performance and energy efficiency of the application. For instance preference learning may enable to compute on-line new priorities to back-fill the ready jobs.

4.1.5 Multi-objective Optimization

Several optimization questions that arise in allocation and scheduling problems lead to the study of several objectives at the same time. The goal is then not a single optimal solution, but a more complicated mathematical object that captures the notion of trade-off. In broader terms, the goal of multi-objective optimization is not to externally arbitrate on disputes between entities with different goals, but rather to explore the possible solutions to highlight the whole range of interesting compromises. A classical tool for studying such multi-objective optimization problems is to use Pareto curves. However, the full description of the Pareto curve can be very hard because of both the number of solutions and the hardness of computing each point. Addressing this problem will opens new methodologies for the analysis of algorithms.

To further illustrate this point here are three possible case studies with emphasis on conflicting interests measured with different objectives. While these cases are good representatives of our HPC context, there are other pertinent trade-offs we may investigate depending on the technology evolution in the coming years. This enumeration is certainly not limitative.

Energy versus Performance. The classical scheduling algorithms designed for the purpose of performance can no longer be used because performance and energy are contradictory objectives to some extent. The scheduling problem with energy becomes a multi-objective problem in nature since the energy consumption should be considered as equally important as performance at exascale. A global constraint on energy could be a first idea for determining trade-offs but the knowledge of the Pareto set (or an approximation of it) is also very useful.

Administrators versus application developers. Both are naturally interested in different objectives: In current algorithms, the performance is mainly computed from the point of view of administrators, but the users should be in the loop since they can give useful information and help to the construction of better schedules. Hence, we face again a multi-objective problem where, as in the above case, the approximation of the Pareto set provides the trade-off between the administrator view and user demands. Moreover, the objectives are usually of the same nature. For example, max stretch and average stretch are two objectives based on the slowdown factor that can interest administrators and users, respectively. In this case the study of the norm of stretch can be also used to describe the trade-off (recall that the $L_1$-norm corresponds to the average objective while the $L_{\infty }$-norm to the max objective). Ideally, we would like to design an algorithm that gives good approximate solutions at the same time for all norms. The $L_2$ or $L_3$-norm are useful since they describe the performance of the whole schedule from the administrator point of view as well as they provide a fairness indication to the users. The hard point here is to derive theoretical analysis for such complicated tools.

In general, resource augmentation can explain the intuitive good behavior of some greedy algorithms while, more interestingly, it can give ideas for new algorithms. For example, in the rejection context we could dedicate a small number of nodes for the usually problematic rejected jobs. Some initial experiments show that this can lead to a schedule for the remaining jobs that is very close to the optimal one.

4.2.1 Workload Traces with Resource Consumption

Workload traces are the base element to capture the behavior of complete systems composed of submitted jobs, running applications, and operating tools. These traces must be obtained on production platforms to provide relevant and representative data. To get a better understanding of the use of such systems, we need to look at both, how the jobs interact with the job management system, and how they use the allocated resources. We propose a general workload trace format that adds jobs resource consumption to the commonly used Standard Workload Format workload trace format. This requires to instrument the platforms, in particular to trace resource consumptions like CPU, data movements at memory, network and I/O levels, with an acceptable performance impact. In a previous work we studied and proposed a dedicated job monitoring tool whose impact on the system has been measured as lightweight (0.35$\%$ speed-down) with a 1 minute sampling rate. Other tools also explore job monitoring, like TACC Stats. A unique feature from our tool is its ability to monitor distinctly jobs sharing common nodes.

Collected workload traces with jobs resource consumption will be publicly released and serve to provide data for works presented in Section 4.1. The trace analysis is expected to give valuable insights to define models encompassing complex behaviours like network topology sensitivity, network congestion and resource interferences.

We expect to join efforts with partners for collecting quality traces (ATOS/Bull, Ciment meso center, Joint Laboratory on Extreme Scale Computing) and will collaborate with the INRIA team POLARIS for their analysis.

4.2.2 Simulation

Simulations of large scale systems are faster by multiple orders of magnitude than real experiments. Unfortunately, replacing experiments with simulations is not as easy as it may sound, as it brings a host of new problems to address in order to ensure that the simulations are closely approximating the execution of typical workloads on real production clusters. Most of these problems are actually not directly related to scheduling algorithms assessment, in the sense that the workload and platform models should be defined independently from the algorithm evaluations, in order to ensure a fair assessment of the algorithms' strengths and weaknesses. These research topics (namely platform modeling, job models and simulator calibration) are addressed in the other subsections.

We developed an open source platform simulator within DataMove (in conjunction with the OAR development team) to provide a widely distributable test bed for reproducible scheduling algorithm evaluation. Our simulator, named Batsim, allows to simulate the behavior of a computational platform executing a workload scheduled by any given scheduling algorithm. To obtain sound simulation results and to broaden the scope of the experiments that can be done thanks to Batsim, we did not chose to create a (necessarily limited) simulator from scratch, but instead to build on top of the SimGrid simulation framework.

To be open to as many batch schedulers as possible, Batsim decouples the platform simulation and the scheduling decisions in two clearly-separated software components communicating through a complete and documented protocol. The Batsim component is in charge of simulating the computational resources behaviour whereas the scheduler component is in charge of taking scheduling decisions. The scheduler component may be both a resource and a job management system. For jobs, scheduling decisions can be to execute a job, to delay its execution or simply to reject it. For resources, other decisions can be taken, for example to change the power state of a machine i.e. to change its speed (in order to lower its energy consumption) or to switch it on or off. This separation of concerns also enables interfacing with potentially any commercial RJMS, as long as the communication protocol with Batsim is implemented. A proof of concept is already available with the OAR RJMS.

Using this test bed opens new research perspectives. It allows to test a large range of platforms and workloads to better understand the real behavior of our algorithms in a production setting. In turn, this opens the possibility to tailor algorithms for a particular platform or application, and to precisely identify the possible shortcomings of the theoretical models used.

4.2.3 Job and Platform Models

The central purpose of the Batsim simulator is to simulate job behaviors on a given target platform under a given resource allocation policy. Depending on the workload, a significant number of jobs are parallel applications with communications and file system accesses. It is not conceivable to simulate individually all these operations for each job on large plaforms with their associated workload due to implied simulation complexity. The challenge is to define a coarse grain job model accurate enough to reproduce parallel application behavior according to the target platform characteristics. We will explore models similar to the BSP (Bulk Synchronous Program) approach that decomposes an application in local computation supersteps ended by global communications and a global synchronization. The model parameters will be established by means of trace analysis as discussed previously, but also by instrumenting some parallel applications to capture communication patterns. This instrumentation will have a significant impact on the concerned application performance, restricting its use to a few applications only. There are a lot of recurrent applications executed on HPC platform, this fact will help to reduce the required number of instrumentations and captures. To assign each job a model, we are considering to adapt the concept of application signatures as proposed in. Platform models and their calibration are also required. Large parts of these models, like those related to network, are provided by Simgrid. Other parts as the filesystem and energy models are comparatively recent and will need to be enhanced or reworked to reflect the HPC platform evolutions. These models are then generally calibrated by running suitable benchmarks.

4.2.4 Emulation and Reproducibility

The use of coarse models in simulation implies to set aside some details. This simplification may hide system behaviors that could impact significantly and negatively the metrics we try to enhance. This issue is particularly relevant when large scale platforms are considered due to the impossibility to run tests at nominal scale on these real platforms. A common approach to circumvent this issue is the use of emulation techniques to reproduce, under certain conditions, the behavior of large platforms on smaller ones. Emulation represents a natural complement to simulation by allowing to execute directly large parts of the actual evaluated software and system, but at the price of larger compute times and a need for more resources. The emulation approach was chosen in to compare two job management systems from workload traces of the CURIE supercomputer (80000 cores). The challenge is to design methods and tools to emulate with sufficient accuracy the platform and the workload (data movement, I/O transfers, communication, applications interference). We will also intend to leverage emulation tools like Distem from the MADYNES team. It is also important to note that the Batsim simulator also uses emulation techniques to support the core scheduling module from actual RJMS. But the integration level is not the same when considering emulation for larger parts of the system (RJMS, compute node, network and filesystem).

Replaying traces implies to prepare and manage complex software stacks including the OS, the resource management system, the distributed filesystem and the applications as well as the tools required to conduct experiments. Preparing these stacks generate specific issues, one of the major one being the support for reproducibility. We propose to further develop the concept of reconstructability to improve experiment reproducibility by capturing the build process of the complete software stack. This approach ensures reproducibility over time better than other ways by keeping all data (original packages, build recipe and Kameleon engine) needed to build the software stack.

In this context, the Grid'5000 (see Sec. 7.2) experimentation infrastructure that gives users the control on the complete software stack is a crucial tool for our research goals. We will pursue our strong implication in this infrastructure.

4.3.1 Programming Model and Software Architecture

In situ creates a tighter loop between the scientist and her/his simulation. As such, an in situ framework needs to be flexible to let the user define and deploy its own set of analysis. A manageable flexibility requires to favor simplicity and understandability, while still enabling an efficient use of parallel resources. Visualization libraries like VTK or Visit, as well as domain specific environments like VMD have initially been developed for traditional post-mortem data analysis. They have been extended to support in situ processing with some simple resource allocation strategies but the level of performance, flexibility and ease of use that is expected requires to rethink new environments. There is a need to develop a middleware and programming environment taking into account in its fundations this specific context of high performance scientific analytics.

Similar needs for new data processing architectures occurred for the emerging area of Big Data Analytics, mainly targeted to web data on cloud-based infrastructures. Google Map/Reduce and its successors like Spark or Stratosphere/Flink have been designed to match the specific context of efficient analytics for large volumes of data produced on the web, on social networks, or generated by business applications. These systems have mainly been developed for cloud infrastructures based on commodity architectures. They do not leverage the specifics of HPC infrastructures. Some preliminary adaptations have been proposed for handling scientific data in a HPC context. However, these approaches do not support in situ processing.

Following the initial development of FlowVR, our middleware for in situ processing, we will pursue our effort to develop a programming environment and software architecture for high performance scientific data analytics. Like FlowVR, the map/reduce tools, as well as the machine learning frameworks like TensorFlow, adopted a dataflow graph for expressing analytics pipe-lines. We are convinced that this dataflow approach is both easy to understand and yet expresses enough concurrency to enable efficient executions. The graph description can be compiled towards lower level representations, a mechanism that is intensively used by Stratosphere/Flink for instance. Existing in situ frameworks inherit from the HPC way of programming with a thiner software stack and a programming model close to the machine. Though this approach enables to program high performance applications, this is usually too low level to enable the scientist to write its analysis pipe-line in a short amount of time. The data model, i.e. the data semantics level accessible at the framework level for error check and optimizations, is also a fundamental aspect of such environments. The key/value store has been adopted by all map/reduce tools. Except in some situations, it cannot be adopted as such for scientific data. Results from numerical simulations are often more structured than web data, associated with acceleration data structures to be processed efficiently. We will investigate data models for scientific data building on existing approaches like Adios or DataSpaces.

4.3.2 Resource Sharing

To alleviate the I/O bottleneck, the in situ paradigm proposes to start processing data as soon as made available by the simulation, while still residing in the memory of the compute node. In situ processings include data compression, indexing, computation of various types of descriptors (1D, 2D, images, etc.). Per se, reducing data output to limit I/O related performance drops or keep the output data size manageable is not new. Scientists have relied on solutions as simple as decreasing the frequency of result savings. In situ processing proposes to move one step further, by providing a full fledged processing framework enabling scientists to more easily and thoroughly manage the available I/O budget.

The most direct way to perform in situ analytics is to inline computations directly in the simulation code. In this case, in situ processing is executed in sequence with the simulation that is suspended meanwhile. Though this approach is direct to implement and does not require complex framework environments, it does not enable to overlap analytics related computations and data movements with the simulation execution, preventing to efficiently use the available resources. Instead of relying on this simple time sharing approach, several works propose to rely on space sharing where one or several cores per node, called helper cores, are dedicated to analytics. The simulation responsibility is simply to handle a copy of the relevant data to the node-local in situ processes, both codes being executed concurrently. This approach often lead to significantly beter performance than in-simulation analytics.

For a better isolation of the simulation and in situ processes, one solution consists in offloading in situ tasks from the simulation nodes towards extra dedicated nodes, usually called staging nodes. These computations are said to be performed in-transit. But this approach may not always be beneficial compared to processing on simulation nodes due to the costs of moving the data from the simulation nodes to the staging nodes.

But today the choice of the resource allocation strategy is mostly ad-hoc and defined by the programmer. We will investigate solutions that enable a cooperative use of the resource between the analytics and the simulation with minimal hints from the programmer. In situ processings inherit from the parallelization scale and data distribution adopted by the simulation, and must execute with minimal perturbations on the simulation execution (whose actual resource usage is difficult to know a priori). We need to develop adapted scheduling strategies that operate at compile and run time. Because analysis are often data intensive, such solutions must take into consideration data movements, a point that classical scheduling strategies designed first for compute intensive applications often overlook. We expect to develop new scheduling strategies relying on the methodologies developed in Sec. 4.1.5. Simulations as well as analysis are iterative processes exposing a strong spatial and temporal coherency that we can take benefit of to anticipate their behavior and then take more relevant resources allocation strategies, possibly based on advanced learning algorithms or as developed in Section 4.1.

In situ analytics represent a specific workload that needs to be scheduled very closely to the simulation, but not necessarily active during the full extent of the simulation execution and that may also require to access data from previous runs (stored in the file system or on specific burst-buffers). Several users may also need to run concurrent analytics pipe-lines on shared data. This departs significantly from the traditional batch scheduling model, motivating the need for a more elastic approach to resource provisioning. These issues will be conjointly addressed with research on batch scheduling policies (Sec. 4.1).

4.3.3 Co-Design with Data Scientists

Given the importance of users in this context, it is of primary importance that in situ tools be co-designed with advanced users, even if such multidisciplinary collaborations are challenging and require constant long term investments to learn and understand the specific practices and expectations of the other domain.

We will tightly collaborate with scientists of some application domains, like molecular dynamics or fluid simulation, to design, develop, deploy and assess in situ analytics scenarios.

7.1.1 OAR

• Keywords:
  HPC, Cloud, Clusters, Resource manager, Light grid
• Scientific Description:
  This batch system is based on a database (PostgreSQL (preferred) or MySQL), a script language (Perl) and an optional scalable administrative tool (e.g. Taktuk). It is composed of modules which interact mainly via the database and are executed as independent programs. Therefore, formally, there is no API, the system interaction is completely defined by the database schema. This approach eases the development of specific modules. Indeed, each module (such as schedulers) may be developed in any language having a database access library.
• Functional Description:
  OAR is a versatile resource and task manager (also called a batch scheduler) for HPC clusters, and other computing infrastructures (like distributed computing experimental testbeds where versatility is a key).
• URL:
  http://­oar.­imag.­fr
• Contact:
  Olivier Richard
• Participants:
  Bruno Bzeznik, Olivier Richard, Pierre Neyron
• Partners:
  LIG, CNRS, Grid'5000, CIMENT, UAR GRICAD

7.1.2 MELISSA

• Keywords:
  Sensitivity Analysis, HPC, Data assimilation, Exascale, AI4Science
• Functional Description:
  Melissa is a middleware framework for on-line processing of data produced from large scale ensemble runs (parameter sweep data analysis) for sensibility analysis, data assimilation and deep surrogate training. Largest runs so far involved up to 30k core, executed 80 000 parallel simulations, and generated 288 TB of intermediate data that did not need to be stored on the file system. For deep surrogate training Melissa demonstrated it can significantly speed-up training on multiple GPUs by maintaining a very high GPU usage.
• URL:
  https://­gitlab.­inria.­fr/­melissa
• Publications:
  hal-04145897, hal-04213978, hal-04102400, hal-01383860, hal-01607479, hal-03017033, hal-03927612, hal-03842106
• Contact:
  Bruno Raffin
• Partner:
  Edf

7.1.3 NixOS-Compose

• Keywords:
  Infrastructure software, Deployment, High performance computing, Distributed computing
• Functional Description:
  NixOS-Compose simplifies the process of setting up ephemeral distributed systems by utilizing Nix's functional package management and NixOS's declarative configuration management. The tool facilitates testing, development, infrastructure prototyping, benchmarking, and advanced experiments in high-performance computing by providing easy and reproducible software stack deployment.
• URL:
  https://­gitlab.­inria.­fr/­nixos-compose/­nixos-compose
• Publication:
  hal-03723771
• Contact:
  Olivier Richard
• Partners:
  LIG, CNRS, UGA

7.1.4 Batsim

• Functional Description:
  BatSim is a Resource and Job Management System (RJMS) framework simulator based on SimGrid. It aims at taking into account platform's hardware capabilities and impacts in simulations. Also, schedulers parts are plugable through a comprehensive API and they are seen as external component of the framework.
• Release Contributions:
  see https://batsim.readthedocs.io/en/latest/changelog.html
• URL:
  https://­batsim.­readthedocs.­io/­en/­latest/
• Contact:
  Olivier Richard

7.1.5 Kameleon

• Keyword:
  Engineering software systems
• Functional Description:
  Kameleon is a simple but powerful tool to generate customized appliances. With Kameleon, you make your recipe that describes how to create step by step your own distribution. At start Kameleon is used to create custom kvm, docker, VirtualBox, ..., but as it is designed to be very generic you can probably do a lot more than that.
• URL:
  http://­kameleon.­imag.­fr/
• Contact:
  Olivier Richard
• Participant:
  Olivier Richard
• Partner:
  Grid'5000

7.1.6 alumet

• Name:
  ALUMET: unified measurement software
• Keywords:
  Energy, Rust, Power monitoring, High performance computing, Performance measure
• Functional Description:
  Alumet provides a generic measurement pipeline with three steps: poll measurement sources, transform the data, and write the result. It is designed to be able to ingest metrics from various sources without redundant work. Supported sources include RAPL domains, Nvidia's NVML, and Jetson INA sensors. The list of supported devices will quickly grow over time, thanks to the next feature of Alumet.
• URL:
  https://­alumet.­dev/
• Contact:
  Guillaume Raffin
• Partner:
  Bull - Atos Technologies

7.2.1 SILECS/Grid'5000 and Meso Center Ciment

We are very active in promoting the factorization of compute resources at a regional and national level. We have a three level implication, locally to maintain a pool of very flexible experimental machines (hundreds of cores), regionally through the GRICAD meso center, and nationally by contributing to the Slices-fr/Grid'5000 platform, our local resources being included in this platform. Olivier Richard is member of Slices-fr/Grid'5000 scientific committee. The OAR scheduler in particular is deployed on both infrastructures. DataMove is hosting several engineers dedicated to Grid'5000 support.

Estimating the environmental impact of Generative-AI services using an LCA-based methodology.

This research addresses the growing environmental concerns related to Generative AI technologies. Using Stable Diffusion as a case study, we developed a life-cycle analysis methodology to calculate the full environmental costs of Gen-AI services from end to end. Our findings reveal that these services generate significant impact through user terminals and networks. Importantly, we demonstrate that merely decarbonizing electricity sources will be insufficient due to consumption of both energy and rare metals. The methodology differentiates between embodied and operational impacts, providing valuable data for transparent discussions about Gen-AI sustainability.

A Methodology and a Toolbox to Explore Dataset related to the Environmental Impact of HTTP Requests.

This work critically analyzes the EcoIndex approach for evaluating environmental performance of URLs. Rather than calculating a plain score, we propose alternative methods that contextualize queries among others. Our contributions include extensive statistical experiments, low-cost Machine Learning approaches, and an implementation available on GitHub. The research raises important questions about relevant data policies for studying digital environmental impacts and promotes synergy between technical expertise and business knowledge.

Light-weight prediction for improving energy consumption in HPC platforms.

This study addresses power-aware resource management in High-performance computing (HPC) systems. Using logs from Marconi 100 (a 980-server supercomputer), we developed a method to predict and exploit workload power consumption with the goal of reducing power usage while maintaining scheduling performance. Our approach combines light-weight power prediction methods with a classical EASY Backfilling inspired heuristic. Simulation results demonstrate that accurate power prediction can improve energy management without significant negative impact on performance.

Dynamic Load/Propagate/Store for Data Assimilation with Particle Filters on Supercomputers.

This paper presents a framework for efficiently running data assimilation methods on supercomputers. Our approach uses an elastic and fault-tolerant runner/server architecture that minimizes data movements while enabling dynamic load balancing. The framework was validated with a bootstrap particle filter using the WRF simulation code, successfully handling 2,555 particles on 20,442 compute cores. Compared to file-based implementations, our solution uses up to 2.84 times fewer resources per particle.

MelissaDL x Breed: Towards Data-Efficient On-line Supervised Training of Multi-parametric Surrogates with Active Learning.

Building on our previous Melissa framework, this research introduces a new active learning method to enhance data-efficiency for on-line surrogate training. Our approach uses Adaptive Multiple Importance Sampling guided by training loss statistics to focus neural network training on difficult areas of the parameter space. Preliminary results for 2D heat PDE demonstrate the potential of this method (called Breed) to improve generalization capabilities while reducing computational overhead.

Extreme-scale workflows: A perspective from the JLESC international community.

This paper presents representative extreme-scale workflows and workflow systems developed by Joint Laboratory for Extreme-Scale Computing (JLESC) participating institutions. We share lessons learned during tool development, alongside open challenges and future research directions in extreme-scale workflows.

Handling Delayed Feedback in Distributed Online Optimization: A Projection-Free Approach.

This study investigates online convex optimization under adversarial delayed feedback. We propose two projection-free algorithms for centralized and distributed settings that achieve optimal regret bounds while remaining projection-free. The research includes theoretical analysis and experimental validation on real-world problems, comparing our algorithms with existing approaches.

Longevity of Artifacts in Leading Parallel and Distributed Systems Conferences: a Review of the State of the Practice in 2023.

This survey examines nearly 300 articles from five leading conferences on parallel and distributed systems held in 2023. We gathered information about artifacts, sharing methods, experimental setups, and reproducibility badges. Our findings indicate that current practices do not adequately address artifact longevity. We propose a new badge based on source code, experimental setup, and software environment to reward artifacts expected to withstand the test of time.

Modeling and solving framework for tactical maintenance planning problems with health index considerations.

This work presents an original tactical maintenance planning problem involving health index indicators for equipment components. We developed a lot-sizing based approach where maintenance is viewed as a form of delayed production. The problem addresses both cost minimization and resource consumption objectives. As this problem is proven to be strongly NP-hard in the general case, we propose a bi-objective Mixed Integer Linear Program and two matheuristics: a degradation matheuristic and a Pareto-based matheuristic.

10.1.1 Horizon Europe

10.1.2 H2020 projects

10.2.1 PEPR NUMPEX

• Goals:
  The main objective of the NumPEx (Numeric for Exascale) program in France is to develop state-of-the-art skills and infrastructures in the field of exascale computing.
• Duration:
  From 2023 to 2030
• Web site:
  NUMPEX
• Datamove implication:
  • Exa-DoST (Data-oriented Software and Tools for the Exascale): Co-lead WP3.
  • Exa-AToW (Architectures and Tools for Large-Scale Workflows): Co-lead WP5.
  • Exa-DI (Development and integration): CO-lead WP3.
• Datamove budget:
  1.295 M euros.

10.2.2 BPI

• Projet AMI Cloud OTPaaS (2021-2024). Aims at offering a new Cloud offer, compatible with Gaia-X and easy to use, that could favour the massive digital transition of companies. Datamove Budget: 110 Keuro.

10.2.3 ANR

• PPR Océan et Climat MEDIATION (2022-2030). Methodological developments for a robust and efficient digital twin of the ocean. Pi: INRIA team AIRSEA. Partners: INRIA, CNRS, IFREMER, IRD, Université Aix-Marseille, Institut National Polytechnique de Toulouse, Ecole Nationale Supérieure Mines-Télécom Atlantique Bretagne Pays de la Loire, Service Hygrodgraphique et Océanographique de la Marine, Université Grenoble Alpes, Météo-France-DESR-Centre National de Recherches Météorologiques. Total budget: 2,4 Meuros. Datamove Budget: 110 Keuros. CO-lead of the WP Leveraging AI and HPC for Digital Twins of the Ocean.
• AAPG2023 PREDICTIONS (2024-2027). This project aims to substantially strengthen and expand the foundations of the nascent, but fast-growing area of algorithms with predictions, in a global framework that addresses all aspects of algorithm development: modeling, design, framework of analysis, and performance evaluation. Specifically, we put forward three main objectives. Pi: LIP6. Partners: LIG/Datamove, IRIF,CC-IN2P3, LIRIS. Total budget: 358k euros.Datamova Budget: 128k euros. Spyros ANGELOPOULOS

11.1.1 Scientific events: organisation

• Co-chair JRAF - Journées de Recherche en Apprentissage Frugal
• Co-chair GAP - Grenoble Artificial Intelligence for Physical Sciences

11.1.2 Invited talks

• "AI and HPC: Trends". September 24th, 2024 ACM's Special Interest Group in High Performance Computing (SIGHPC) Computing Continuum Chapter.
• "Online Training from Numerical Simulations" at WANT Want workshop at ICML 204, Vienna, 2024.
• “Entender e medir o consumo energético de aplicações de Inteligência Artificial”. VII Escola Regional de Alto Desempenho do Centro-Oeste, Brasilia, Brazil

11.1.3 Leadership within the scientific community

• Steering committee member of GDR-RSD (Réseaux et Systèmes Distribués)

11.1.4 Scientific expertise

• DAVID Lab HCERES committee member.

International journals

• 7 articleA.Adrien Berthelot, E.Eddy Caron, M.Mathilde Jay and L.Laurent Lefevre. Understanding the environmental impact of generative AI services.Communications of the ACMSpecial Issue on Sustainability and Computing2025, 1-14In press. HAL
• 8 articleR.Robin Boëzennec, F.Fanny Dufossé and G.Guillaume Pallez. Qualitatively Analyzing Optimization Objectives in the Design of HPC Resource Manager.ACM Transactions on Modeling and Performance Evaluation of Computing Systems2024, 1-28In press. HAL
• 9 articleE.Ernest Foussard, M.Margaux Nattaf, M.-L.Marie-Laure Espinouse and G.Grégory Mounié. Modeling and solving framework for tactical maintenance planning problems with health index considerations.Computers and Operations Research170October 2024, 106763HALDOI
• 10 articleS.Sebastian Friedemann, K.Kai Keller, Y.-S.Yen-Sen Lu, B.Bruno Raffin and L.Leonardo Bautista Gomez. Dynamic Load/Propagate/Store for Data Assimilation with Particle Filters on Supercomputers.Journal of computational scienceJanuary 2024, 102229HALDOI
• 11 articleG.Guillaume Raffin and D.Denis Trystram. Dissecting the software-based measurement of CPU energy consumption: a comparative analysis.IEEE Transactions on Parallel and Distributed Systems361November 2024, 96HALDOI
• 12 articleO.Orcun Yildiz, A.Amal Gueroudji, J.Julien Bigot, B.Bruno Raffin, R. M.Rosa M. Badia and T.Tom Peterka. Extreme-scale workflows: A perspective from the JLESC international community.Future Generation Computer Systems1612024, 502-513HALDOI

International peer-reviewed conferences

• 13 inproceedingsA.Adrien Berthelot, E.Eddy Caron, M.Mathilde Jay and L.Laurent Lefèvre. Estimating the environmental impact of Generative-AI services using an LCA-based methodology.Procedia CIRPCIRP LCE 2024 - 31st Conference on Life Cycle EngineeringTurin, Italy2024, 1-10HAL
• 14 inproceedingsR.Robin Boëzennec, D.Danilo Carastan-Santos, F.Fanny Dufossé and G.Guillaume Pallez. Allocation Strategies for Disaggregated Memory in HPC Systems.HiPC 2024 - 31st IEEE International Conference on High Performance Computing, Data, and AnalyticsBengalore, IndiaIEEE2024, 1-11HAL
• 15 inproceedingsD.Danilo Carastan-Santos, G.Georges da Costa, M.Millian Poquet, P.Patricia Stolf and D.Denis Trystram. Light-weight prediction for improving energy consumption in HPC platforms.Lecture Notes in Computer ScienceEuro-Par 202414801Madrid, SpainSpringerAugust 2024, 152-165HALDOI
• 16 inproceedingsC.Christophe Cérin, M.Mathilde Jay and L.Laurent Lefèvre. A Methodology and a Toolbox to Explore Dataset related to the Environmental Impact of HTTP Requests.2023 IEEE International Conference on Big Data (BigData) - 3rd International Workshop on Big Data Analytics for SustainabilitySorrento (Naples), ItalyIEEE2024, 1-10HALDOI
• 17 inproceedingsC.Christophe Cérin, M.Mamadou Sow, D.Didier Donsez and L. C.Louis Closson Adeunis. Towards Online Machine Learning Libraries for Embedded Systems.WF-IoT 2024 - IEEE 10th World Forum on Internet of ThingsOttawa, CanadaIEEE2024, 790-797HALDOI
• 18 inproceedingsL.Liticia Djennadi, G.Gladys Diaz, K.Khaled Boussetta and C.Christophe Cerin. SDN-based approach for adaptive reconfiguration of routing in IoT for smart-buildings.HPSH 2024 - 25th IEEE International Conference on High Performance Switching and RoutingPisa, ItalyIEEEAugust 2024, 137-142HALDOI
• 19 inproceedingsP.-F.Pierre-François Dutot. REGALE project: Oar, A Versatile Resource and JobManagement System.CONCERTO 2024 - 2nd workshop on projeCts crOss-synergy iN advanCing Exascale platfoRms and quanTum cOmputingMunich, Germany2024HAL
• 20 inproceedingsS.Sofya Dymchenko, A.Abhishek Purandare and B.Bruno Raffin. MelissaDL x Breed: Towards Data-Efficient On-line Supervised Training of Multi-parametric Surrogates with Active Learning.SC-W 2024 - Workshops of The International Conference on High Performance Computing, Network, Storage, and AnalysisAI4S 2024 - 5th Workshop on artificial intelligence and machine learning for scientific applicationsAtlanta (Georgia), United StatesIEEE2024, 1-9HAL
• 21 inproceedingsQ.Quentin Guilloteau, F. M.Florina M Ciorba, M.Millian Poquet, D.Dorian Goepp and O.Olivier Richard. Longevity of Artifacts in Leading Parallel and Distributed Systems Conferences: a Review of the State of the Practice in 2023.REP 2024 - ACM Conference on Reproducibility and ReplicabilityRennes, France2024, 1-14HALDOI
• 22 inproceedingsT.Tarek Menouer, C.Christophe Cérin and P.Patrice Darmon. Reactive Autoscaling of Kubernetes Nodes.FRAME 2024 - 4th Workshop on Flexible Resource and Application Management on the EdgePisa, ItalyACM2024, 31 - 38HAL
• 23 inproceedingsT.-A.Tuan-Anh Nguyen, N.Nguyen Kim Thang and D.Denis Trystram. Handling Delayed Feedback in Distributed Online Optimization: A Projection-Free Approach.Lecture Notes in Artificial IntelligenceECML PKDD 2024 - Joint European Conference on Machine Learning and Knowledge Discovery in Databases14941Lecture Notes in Computer ScienceVilnius, LithuaniaSpringer Nature SwitzerlandAugust 2024, 197-211HALDOI

National peer-reviewed Conferences

• 24 inproceedingsY.Yoann Dupas, O.Olivier Hotel, G.Grégoire Lefebvre and C.Christophe Cérin. Empreinte Énergétique de YoloV8 pour la Détection de Piétons et de Véhicules.Revue des Nouvelles Technologies de l'Information,EGC 2024 - 24ème conférence francophone sur l'Extraction et la Gestion des ConnaissancesRNTI-E-40Dijon, FranceJanuary 2024, 345-346HAL

Conferences without proceedings

• 25 inproceedingsD.Dorian Goepp, S.Samuel Brun, Q.Quentin Guilloteau and O.Olivier Richard. Un prototype de cache de métadonnées pour le passage à l'échelle de NixOS-Compose.COMPAS 2024 - Conférence francophone d'informatique en Parallélisme, Architecture et SystèmeNantes, France2024, 1-8HAL
• 26 inproceedingsG.Guillaume Schreiner and P.Pierre Neyron. SLICES-FR : l’infrastructure de recherche nationale pour l’expérimentation Cloud et Réseaux du futur.JRES 2024 - Journées réseaux de l'enseignement et de la rechercheRennes, FranceDecember 2024, 1-15HAL
• 27 inproceedingsM.Mamadou Sow and M.Muhammad Sanaullah Kayani. Online Machine Learning for Embedded Systems (ESP32).ComPAS 2024 - Conférence francophone en informatique autour des thématiques du parallélisme, de l'architecture et des systèmesNantes, France2024, 1-20HAL

Scientific book chapters

• 28 inbookL.Louis Closson, C.Christophe Cérin, D.Didier Donsez and J.-L.Jean-Luc Baudouin. Design of a Meaningful Framework for Time Series Forecasting in Smart Buildings.15MDPI InformationFebruary 2024, 94HALDOI
• 29 inbookI.Imad Kissami, C.Christophe Cérin, F.Fayssal Benkhaldoun and F.Fahd Kalloubi. Facts and issues of neural networks for numerical simulation.Artificial Intelligence and High-Performance Computing in the Cloud - Research and Application Challenges.Lecture Notes in Networks and Systems2024. In press. HAL

Edition (books, proceedings, special issue of a journal)

• 30 proceedingsEco-responsible Digital Day.Journée Numérique Écoresponsable à VilletaneuseOctober 2024HAL

Doctoral dissertations and habilitation theses

• 31 thesisM.Mathilde Jay. A Versatile Methodology for Assessing the Electricity Consumption and Environmental Footprint of Machine Learning Training: from Supercomputers to Edge Devices.Université Grenoble AlpesOctober 2024HAL
• 32 thesisL.Lucas Meyer. Online Deep Learning for Numerical Simulations at Scale.Université Grenoble Alpes [2020-....]March 2024HAL
• 33 thesisT.-A.Tuan-Anh Nguyen Huu. Online Distributed Learning : A Projection-Free Approach.Université Grenoble Alpes [2020-....]October 2024HAL

Reports & preprints

• 34 miscL.Luc Angelelli, D.Danilo Carastan-Santos and P.-F.Pierre-François Dutot. Run your HPC jobs in Eco-Mode: revealing the potential of user-assisted power capping in supercomputing systems.March 2024HAL
• 35 miscM.Mathieu Bacou, D.David Beserra, E.Eugen Dedu, L.Loïc Desgeorges, D.Didier Donsez, A.Alexandre Guitton, B.Baptiste Jonglez, A.Arnaud Legrand, G.Georgios Papadopoulos, O.Olivier Richard, S.Samir Si-Mohammed, N.Nina Tamdrari and F.Fabrice Theoleyre. Journée thématique du GDR RSD : pratiques expérimentales de la communauté systèmes et réseaux.January 2025HAL
• 36 miscA.Abdessalam Benhari, Y.Yves Denneulin, F.Frédéric Desprez, F.Fanny Dufossé and D.Denis Trystram. Green HPC: An analysis of the domain based on Top500.March 2024HAL
• 37 reportS.Sylvain Bouveret, A.Aurélie Bugeau, F.Frenoux Emmanuelle, J.Julien Lefevre, L.Laurent Lefèvre, A.-L.Anne-Laure Ligozat, P.Philippe Marquet, A.-C.Anne-Cécile Orgerie and D.Denis Trystram. Quiz sur les impacts environnementaux du numérique.EcoInfoFebruary 2025, 1-5HAL
• 38 miscD.Danilo Carastan-Santos, G.Georges da Costa, I.Igor Fontana de Nardin, M.Millian Poquet, K.Krzysztof Rzadca, P.Patricia Stolf and D.Denis Trystram. Scheduling with lightweight predictions in power-constrained HPC platforms.October 2024HAL
• 39 reportA.Alban Dutilleul, H.Hugo Pompougnac, N.Nicolas Derumigny, G.Gabriel Rodríguez, V.Valentin Trophime, C.Christophe Guillon and F.Fabrice Rastello. Performance debugging through microarchitectural sensitivity and causality analysis.INRIADecember 2024, 1-13HALDOI
• 40 reportA.Alena Shilova, T.Thomas Delliaux, P.Philippe Preux and B.Bruno Raffin. Learning HJB Viscosity Solutions with PINNs for Continuous-Time Reinforcement Learning.RR-9541Inria Lille - Nord Europe, CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189; Univ. Lille, CNRS, Centrale Lille, Inria UMR 9189 - CRIStAL,INRIA Lille Nord Europe, Villeneuve d’Ascq, France; Univ. Grenoble Alps, CNRS, Inria, Grenoble INP, LIG, 38000 Grenoble, FranceFebruary 2024, 1-30HAL

Other scientific publications

• 41 inproceedingsA.Adrien Berthelot, E.Eddy Caron, M.Mathilde Jay and L.Laurent Lefèvre. Towards a multi-criteria evaluation of the environmental footprint of generative ai services.ICT4S 2024 - International Conference on Information and Communications Technology for SustainabilityStockholm, Sweden2024, 1-1HAL
• 42 miscD.Dorian Goepp, F. A.Fernando Ayats Llamas, O.Olivier Richard and Q.Quentin Guilloteau. ACM REP24 Tutorial: Reproducible distributed environments with NixOS Compose.June 2024, 1-3HAL

Scientific popularization

• 43 article T.Thierry Ménissier and D.Denis Trystram. Can AI really be frugal? The Conversation France May 2024 HAL
• 44 miscP.Pierre Neyron. R’n’R building of an experimentation platform for the Edge.Chichillianne, FranceJune 2024, 1-12HAL

Software

• 45 softwareO.Olivier Richard. OAR2: Versatile Resource and Job Manager.November 2024 lic: GNU Library General Public License v2 only.HALSoftware HeritageVCS