Energy-aware actor-based distributed programming

Domain and scientific/technical context

The energy consumption of software infrastructures and applications, notably in the do- mains of the Cloud-Edge-IoT (CEI) continuum and AI-intensive software systems is a major challenge for today’s societies. Controlling the energy consumption is notably crucial in two different contexts: (i) the management/optimization of the overall energy consumption of large-scale software systems among with AI-intensive software systems now occupy a place of choice and (ii) the integration of massive numbers and large varieties of small, battery powered IoT devices in large-scale distributed systems.

Currently, energy consumption is typically handled by runtime monitoring of compo- nents of the software systems running on distributed systems, e.g., data centers, grid ar- chitectures, sensor networks and other distributed cyber-physical infrastructures. Based on measured energy consumption and (configured) available energy quotas, some scheduling component then dynamically enables computations based on energy policies. [SDU+23, HCW+21, CKC+21]

Scientific/technical challenges

Current distributed software/hardware systems almost never provide guarantees that future computations (that are typically encapsulated in some software component) can be executed given the currently available energy. Furthermore, energy-related information is currently almost always expressed at the API and systems-level [RWS+24]: energy-related properties therefore cannot be expressed at the programming level. Such guarantees require a precise link between energy-consuming computations, i.e., distributed programs, energy sources (available electricity networks, batteries) and available energy quotas determined based through dynamic monitoring.

Such a link has been proposed for non-distributed programs (through program language extensions or transformations [CY17, ZLL15, YVS17] as well as static analysis [MSP+21, KKK16]) but is lacking for distributed systems. Currently, software systems for distributed systems are typically structured in terms of separate programs that are deployed on different machines, often using containers [SDU+23], and cooperate at runtime. More precise energy policies and energy-aware schedulers can thus be defined only in terms of specifications that are difficult to link to the set of executing programs.

Considered methods, targeted results and impacts

The main goal of this PhD is the precise definition and efficient enforcement of energy contracts over distributed programs. We are targeting a distributed pro- gramming language extension that allows required and available energy quotas of programs to be partially defined by the programmer and partially provided dynamically by the dis- tributed environment in which the program is executed. Required energy quotas are then checked using static analysis or dynamic tests against the energy available from the execu- tion environment. Energy quotas are managed and enforced through contexts that energy- aware programs use to store energy-related environmental information. Contexts are also harnessed in order to actively manage energy-consuming computations using placement of computations, optimization of communication, rate limiting, scaling etc.

Distributed computations are represented and energy contexts maintained in the envi- sioned programming method by means of software actors [MSD18, HSF16], components that encapsulate computations and data that can be flexibly deployed and executed in distributed systems.

Environment (partners, places, specific tools and hardware)

At IMT Atlantique, the PhD student will be part of the STACK team, a leading team on research on distributed programming and infrastructures, notably in the Cloud-Edge-IoT continuum.

At the VUB, the PhD student will be part of the Soft laboratory, a leading lab in the domain of programming languages.

Interdisciplinarity aspects

The topic straddles the domains of software engineering, energy efficiency and frugal com- puting that is useful in numerous other domains. It includes strong interdisplinary aspects, notably through applications in the fields of data science (e.g., energy optimization of data

analyses) and the Industry of the Future (e.g., the optimization of monolithic batch pro- cesses that are currently used frequently). These interdisciplinary aspects will be explored, in particular, together with the non academic partners of this topic.

Supervisors and study periods

•

•

•

2.2

2.2.1

[MSD18] Myter, F., Scholliers, C., De Meuter, W. (2018). Parallel and Distributed Web Programming with Actors. In: Ricci, A., Haller, P. (eds) Programming with Actors. Lecture Notes in Computer Science(), vol 10789. Springer, Cham. https: //doi.org/10.1007/978-3-030-00302-9_1

[HSF16] Hayduk, Y., Sobe, A., Felber, P. (2016). Enhanced Energy Efficiency with the Actor Model on Heterogeneous Architectures. In: Jelasity, M., Kalyvianaki, E. (eds) Distributed Applications and Interoperable Systems. DAIS 2016. Lecture Notes in Computer Science(), vol 9687. Springer, Cham. https://doi.org/10. 1007/978-3-319-39577-7_1

Supervisors and study periods

• IMT Atlantique: Prof. Mario Südholt, IMT Atlantique, Nantes, France
  The PhD student will stay 2 years at IMT Atlantique.
• International partner: Prof. Coen De Roover and Prof. Wolfgang De Meuter, Vrije Universiteit Brussel, Brussels, Belgium
  The PhD student will stay 1 year at VUB.
• Industrial partner(s) for short-term visits have not yet been determined. However, cooperations with non-academic partners on similar topics will be harnessed: e.g., with Nantes university hospital for healthcare applications and Zensor SA in Brussels for distributed health monitoring of industrial assets

References

[CY17] Anthony Canino and Yu David Liu. 2017. Proactive and adaptive energy- aware programming with mixed typechecking. In Proceedings of the 38th ACM SIG- PLAN Conference on Programming Language Design and Implementation (PLDI 2017). Association for Computing Machinery, New York, NY, USA, 217–232. https: //doi.org/10.1145/3062341.3062356

[ZLL15] Haitao Steve Zhu, Chaoren Lin, and Yu David Liu. 2015. A programming model for sustainable software. In Proceedings of the 37th International Conference on Software Engineering - Volume 1 (ICSE ’15). IEEE Press, 767–777.

[KKK16] C. H. P. Kim, D. Kroening and M. Kwiatkowska, "Static Program Anal- ysis for Identifying Energy Bugs in Graphics-Intensive Mobile Apps," 2016 IEEE 24th International Symposium on Modeling, Analysis and Simulation of Computer and Telecommunication Systems (MASCOTS), London, UK, 2016, pp. 115-124, doi: 10.1109/MASCOTS.2016.28.

[MSP+21] Marantos, C., Salapas, K., Papadopoulos, L. et al. A Flexible Tool for Esti- mating Applications Performance and Energy Consumption Through Static Analysis. SN COMPUT. SCI. 2, 21 (2021). https://doi.org/10.1007/s42979-020-00405-7

[SDU+23] R. C. Sofia, D. Dykeman, P. Urbanetz, A. Galal and D. A. Dave, "Dynamic, Context-Aware Cross-Layer Orchestration of Containerized Applications," in IEEE Access, vol. 11, pp. 93129-93150, 2023, doi: 10.1109/ACCESS.2023.3307026.

[HCW+21] Yongsheng Hao, Jie Cao, Qi Wang, Jinglin Du, Energy-aware scheduling in edge computing with a clustering method, Future Generation Computer Systems, Volume 117, 2021, Pages 259-272, ISSN 0167-739X, https://doi.org/10.1016/j. future.2020.11.029.

[CKC+21] Chaurasia, N., Kumar, M., Chaudhry, R. et al. Comprehensive survey on energy-aware server consolidation techniques in cloud computing. J Supercomput 77, 11682–11737 (2021). https://doi.org/10.1007/s11227-021-03760-1

[RWS+24] Saurabhsingh Rajput, Tim Widmayer, Ziyuan Shang, Maria Kechagia, Federica Sarro, and Tushar Sharma. 2024. Enhancing Energy-Awareness in Deep Learning through Fine-Grained Energy Measurement. ACM Trans. Softw. Eng. Methodol. 33, 8, Article 211 (November 2024), 34 pages. https://doi.org/10. 1145/3680470

[YVS17] T. -J. Yang, Y. -H. Chen and V. Sze, "Designing Energy-Efficient Convo- lutional Neural Networks Using Energy-Aware Pruning," 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, USA, 2017, pp. 6071-6079, doi: 10.1109/CVPR.2017.643.

[MSD18] Myter, F., Scholliers, C., De Meuter, W. (2018). Parallel and Distributed Web Programming with Actors. In: Ricci, A., Haller, P. (eds) Programming with Actors. Lecture Notes in Computer Science(), vol 10789. Springer, Cham. https: //doi.org/10.1007/978-3-030-00302-9_1

[HSF16] Hayduk, Y., Sobe, A., Felber, P. (2016). Enhanced Energy Efficiency with the Actor Model on Heterogeneous Architectures. In: Jelasity, M., Kalyvianaki, E. (eds) Distributed Applications and Interoperable Systems. DAIS 2016. Lecture Notes in Computer Science(), vol 9687. Springer, Cham. https://doi.org/10. 1007/978-3-319-39577-7_1

IMT Atlantique, internationally recognized for the quality of its research, is a leading French technological university under the supervision of the Ministry of Industry and Digital Technology. IMT Atlantique maintains privileged relationships with major national and international industrial partners, as well as with a dense network of SMEs, start-ups, and innovation networks. With 290 permanent staff, 2,200 students, including 300 doctoral students, IMT Atlantique produces 1,000 publications each year and raises 18€ million in research funds.

Solid knowledge in distributed systems and distributed programming