\n \n \n
\n
\n\n \n \n \n \n \n \n Serverless Cloud Computing: State of the Art and Challenges.\n \n \n \n \n\n\n \n Lannurien, V.; D’Orazio, L.; Barais, O.; and Boukhobza, J.\n\n\n \n\n\n\n In Krishnamurthi, R.; Kumar, A.; Gill, S. S.; and Buyya, R., editor(s),
Serverless Computing: Principles and Paradigms, volume 162, pages 275–316. Springer International Publishing, Cham, 2023.\n
Series Title: Lecture Notes on Data Engineering and Communications Technologies\n\n
\n\n
\n\n
\n\n \n \n ![\"Serverless](\"//bibbase.org/img/filetypes/link.svg\"\n\t "\\"Paper\\"\n\t")
Paper\n \n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@incollection{krishnamurthi_serverless_2023,\n\taddress = {Cham},\n\ttitle = {Serverless {Cloud} {Computing}: {State} of the {Art} and {Challenges}},\n\tvolume = {162},\n\tcopyright = {All rights reserved},\n\tisbn = {978-3-031-26632-4 978-3-031-26633-1},\n\tshorttitle = {Serverless {Cloud} {Computing}},\n\turl = {https://link.springer.com/10.1007/978-3-031-26633-1_11},\n\tabstract = {The serverless model represents a paradigm shift in the cloud: as opposed to traditional cloud computing service models, serverless customers do not reserve hardware resources. The execution of their code is event-driven (HTTP requests, cron jobs, etc.) and billing is based on actual resource usage. In return, the responsibility of resource allocation and task placement lies on the provider. While serverless in the wild is mainly advertised as a public cloud offering, solutions are actively developed and backed by solid actors in the industry to allow the development of private cloud serverless platforms. The first generation of serverless offerings, ”Function as a Service” (FaaS), has severe shortcomings that can offset the potential benefits for both customers and providers – in terms of spendings and reliability on the customer side, and in terms of resources multiplexing on the provider side. Circumventing these flaws would allow considerable savings in money and energy for both providers and tenants. This chapter aims at establishing a comprehensive tour of these limitations, and presenting state-of-the-art studies to mitigate weaknesses that are currently holding serverless back from becoming the de facto cloud computing model. The main challenges related to the deployment of such a cloud platform are discussed and some perspectives for future directions in research are given.},\n\tlanguage = {en},\n\turldate = {2024-06-19},\n\tbooktitle = {Serverless {Computing}: {Principles} and {Paradigms}},\n\tpublisher = {Springer International Publishing},\n\tauthor = {Lannurien, Vincent and D’Orazio, Laurent and Barais, Olivier and Boukhobza, Jalil},\n\teditor = {Krishnamurthi, Rajalakshmi and Kumar, Adarsh and Gill, Sukhpal Singh and Buyya, Rajkumar},\n\tyear = {2023},\n\tdoi = {10.1007/978-3-031-26633-1_11},\n\tnote = {Series Title: Lecture Notes on Data Engineering and Communications Technologies},\n\tpages = {275--316},\n}\n\n\n\n\n
\n\n\n
\n The serverless model represents a paradigm shift in the cloud: as opposed to traditional cloud computing service models, serverless customers do not reserve hardware resources. The execution of their code is event-driven (HTTP requests, cron jobs, etc.) and billing is based on actual resource usage. In return, the responsibility of resource allocation and task placement lies on the provider. While serverless in the wild is mainly advertised as a public cloud offering, solutions are actively developed and backed by solid actors in the industry to allow the development of private cloud serverless platforms. The first generation of serverless offerings, ”Function as a Service” (FaaS), has severe shortcomings that can offset the potential benefits for both customers and providers – in terms of spendings and reliability on the customer side, and in terms of resources multiplexing on the provider side. Circumventing these flaws would allow considerable savings in money and energy for both providers and tenants. This chapter aims at establishing a comprehensive tour of these limitations, and presenting state-of-the-art studies to mitigate weaknesses that are currently holding serverless back from becoming the de facto cloud computing model. The main challenges related to the deployment of such a cloud platform are discussed and some perspectives for future directions in research are given.\n
\n\n\n
\n\n\n
\n
\n\n \n \n \n \n \n \n HeROfake: Heterogeneous Resources Orchestration in a Serverless Cloud – An Application to Deepfake Detection.\n \n \n \n \n\n\n \n Lannurien, V.; D'Orazio, L.; Barais, O.; Bernard, E.; Weppe, O.; Beaulieu, L.; Kacete, A.; Paquelet, S.; and Boukhobza, J.\n\n\n \n\n\n\n In
2023 IEEE/ACM 23rd International Symposium on Cluster, Cloud and Internet Computing (CCGrid), pages 154–165, Bangalore, India, May 2023. IEEE\n
\n\n
\n\n
\n\n
\n\n \n \n Paper\n \n \n\n \n \n doi\n \n \n\n \n link\n \n \n\n bibtex\n \n\n \n \n \n abstract \n \n\n \n\n \n \n \n \n \n \n \n\n \n \n \n\n\n\n
\n
@inproceedings{lannurien_herofake:_2023,\n\taddress = {Bangalore, India},\n\ttitle = {{HeROfake}: {Heterogeneous} {Resources} {Orchestration} in a {Serverless} {Cloud} – {An} {Application} to {Deepfake} {Detection}},\n\tcopyright = {https://doi.org/10.15223/policy-029},\n\tisbn = {9798350301199},\n\tshorttitle = {{HeROfake}},\n\turl = {https://ieeexplore.ieee.org/document/10171518/},\n\tdoi = {10.1109/CCGrid57682.2023.00024},\n\tabstract = {Serverless is a trending service model for cloud computing. It shifts a lot of the complexity from customers to service providers. However, current serverless platforms mostly consider the provider’s infrastructure as homogeneous, as well as the users’ requests. This limits possibilities for the provider to leverage heterogeneity in their infrastructure to improve function response time and reduce energy consumption. We propose a heterogeneity-aware serverless orchestrator for private clouds that consists of two components: the autoscaler allocates heterogeneous hardware resources (CPUs, GPUs, FPGAs) for function replicas, while the scheduler maps function executions to these replicas. Our objective is to guarantee function response time, while enabling the provider to reduce resource usage and energy consumption. This work considers a case study for a deepfake detection application relying on CNN inference. We devised a simulation environment that implements our model and a baseline Knative orchestrator, and evaluated both policies with regard to consolidation of tasks, energy consumption and SLA penalties. Experimental results show that our platform yields substantial gains for all those metrics, with an average of 35\\% less energy consumed for function executions while consolidating tasks on less than 40\\% of the infrastructure’s nodes, and more than 60\\% less SLA violations.},\n\tlanguage = {en},\n\turldate = {2024-06-19},\n\tbooktitle = {2023 {IEEE}/{ACM} 23rd {International} {Symposium} on {Cluster}, {Cloud} and {Internet} {Computing} ({CCGrid})},\n\tpublisher = {IEEE},\n\tauthor = {Lannurien, Vincent and D'Orazio, Laurent and Barais, Olivier and Bernard, Esther and Weppe, Olivier and Beaulieu, Laurent and Kacete, Amine and Paquelet, Stéphane and Boukhobza, Jalil},\n\tmonth = may,\n\tyear = {2023},\n\tpages = {154--165},\n}\n\n
\n\n\n
\n Serverless is a trending service model for cloud computing. It shifts a lot of the complexity from customers to service providers. However, current serverless platforms mostly consider the provider’s infrastructure as homogeneous, as well as the users’ requests. This limits possibilities for the provider to leverage heterogeneity in their infrastructure to improve function response time and reduce energy consumption. We propose a heterogeneity-aware serverless orchestrator for private clouds that consists of two components: the autoscaler allocates heterogeneous hardware resources (CPUs, GPUs, FPGAs) for function replicas, while the scheduler maps function executions to these replicas. Our objective is to guarantee function response time, while enabling the provider to reduce resource usage and energy consumption. This work considers a case study for a deepfake detection application relying on CNN inference. We devised a simulation environment that implements our model and a baseline Knative orchestrator, and evaluated both policies with regard to consolidation of tasks, energy consumption and SLA penalties. Experimental results show that our platform yields substantial gains for all those metrics, with an average of 35% less energy consumed for function executions while consolidating tasks on less than 40% of the infrastructure’s nodes, and more than 60% less SLA violations.\n
\n\n\n
\n\n\n\n\n\n
![\"Loading..\"](\"//bibbase.org/img/ajax-loader.gif\")
\n\n![\"bibbase.org\"\n](\"//bibbase.org/img/logo.svg\"\n "\\"BibBase")
\n \n