OFC: an opportunistic caching system for FaaS platforms
Djob Mvondo
(1, 2)
,
Mathieu Bacou
(3, 4, 5, 6)
,
Kevin Nguetchouang
(7)
,
Lucien Ndjie
(7)
,
Stéphane Pouget
(2)
,
Josiane Kouam
(8)
,
Renaud Lachaize
(1)
,
Jinho Hwang
(9)
,
Tim Wood
(10)
,
Daniel Hagimont
(11, 12)
,
Noël de Palma
(1)
,
Bernabé Batchakui
(7)
,
Alain Tchana
(2, 7)
1
ERODS -
Efficient and Robust Distributed Systems
2 AVALON - Algorithms and Software Architectures for Distributed and HPC Platforms
3 TSP - INF - Département Informatique
4 IP Paris - Institut Polytechnique de Paris
5 ACMES-SAMOVAR - Architecture, Cloud continuum, formal Models, artificial intElligence and Services in distributed computing
6 Benagil-Samovar - Efficient and safe distributed systems
7 ENSPY - Ecole Nationale Supérieure Polytechnique de Yaoundé
8 TRiBE - inTeRnet BEyond the usual
9 Facebook
10 GW - The George Washington University
11 IRIT-SEPIA - Système d’exploitation, systèmes répartis, de l’intergiciel à l’architecture
12 Toulouse INP - Institut National Polytechnique (Toulouse)
2 AVALON - Algorithms and Software Architectures for Distributed and HPC Platforms
3 TSP - INF - Département Informatique
4 IP Paris - Institut Polytechnique de Paris
5 ACMES-SAMOVAR - Architecture, Cloud continuum, formal Models, artificial intElligence and Services in distributed computing
6 Benagil-Samovar - Efficient and safe distributed systems
7 ENSPY - Ecole Nationale Supérieure Polytechnique de Yaoundé
8 TRiBE - inTeRnet BEyond the usual
9 Facebook
10 GW - The George Washington University
11 IRIT-SEPIA - Système d’exploitation, systèmes répartis, de l’intergiciel à l’architecture
12 Toulouse INP - Institut National Polytechnique (Toulouse)
Djob Mvondo
- Fonction : Auteur
- PersonId : 1312307
- IdHAL : djob-mvondo-djob
- ORCID : 0000-0002-2622-7970
Mathieu Bacou
- Fonction : Auteur
- PersonId : 739324
- IdHAL : mathieu-bacou
- ORCID : 0000-0002-1658-9804
- IdRef : 245178384
Lucien Ndjie
- Fonction : Auteur
- PersonId : 1242091
- IdHAL : lndjie
Stéphane Pouget
- Fonction : Auteur
- PersonId : 747693
- IdHAL : stephane-pouget
- ORCID : 0000-0003-3950-5818
Josiane Kouam
- Fonction : Auteur
- PersonId : 1162659
- IdHAL : anne-josiane-kouam
Daniel Hagimont
- Fonction : Auteur
- PersonId : 737290
- IdHAL : daniel-hagimont
- ORCID : 0000-0002-0978-2155
- IdRef : 075122405
Noël de Palma
- Fonction : Auteur
- PersonId : 172307
- IdHAL : noel-de-palma
- ORCID : 0000-0003-4638-7266
- IdRef : 060315555
Alain Tchana
- Fonction : Auteur
- PersonId : 972887
- ORCID : 0000-0001-7789-8400
- IdRef : 15715002X
Résumé
Cloud applications based on the "Functions as a Service" (FaaS) paradigm have become very popular. Yet, due to their stateless nature, they must frequently interact with an external data store, which limits their performance. To mitigate this issue, we introduce OFC, a transparent, vertically and horizontally elastic in-memory caching system for FaaS platforms, distributed over the worker nodes. OFC provides these benefits cost-effectively by exploiting two common sources of resource waste: (i) most cloud tenants overprovision the memory resources reserved for their functions because their footprint is non-trivially input-dependent and (ii) FaaS providers keep function sandboxes alive for several minutes to avoid cold starts. Using machine learning models adjusted for typical function input data categories (e.g., multimedia formats), OFC estimates the actual memory resources required by each function invocation and hoards the remaining capacity to feed the cache. We build our OFC prototype based on enhancements to the OpenWhisk FaaS platform, the Swift persistent object store, and the RAM-Cloud in-memory store. Using a diverse set of workloads, we show that OFC improves by up to 82 % and 60 % respectively the execution time of single-stage and pipelined functions.
Domaines
Système d'exploitation [cs.OS]Format du dépôt | Fichier |
---|---|
Type de dépôt | Communication dans un congrès |
Résumé |
en
Cloud applications based on the "Functions as a Service" (FaaS) paradigm have become very popular. Yet, due to their stateless nature, they must frequently interact with an external data store, which limits their performance. To mitigate this issue, we introduce OFC, a transparent, vertically and horizontally elastic in-memory caching system for FaaS platforms, distributed over the worker nodes. OFC provides these benefits cost-effectively by exploiting two common sources of resource waste: (i) most cloud tenants overprovision the memory resources reserved for their functions because their footprint is non-trivially input-dependent and (ii) FaaS providers keep function sandboxes alive for several minutes to avoid cold starts. Using machine learning models adjusted for typical function input data categories (e.g., multimedia formats), OFC estimates the actual memory resources required by each function invocation and hoards the remaining capacity to feed the cache. We build our OFC prototype based on enhancements to the OpenWhisk FaaS platform, the Swift persistent object store, and the RAM-Cloud in-memory store. Using a diverse set of workloads, we show that OFC improves by up to 82 % and 60 % respectively the execution time of single-stage and pipelined functions.
|
Titre |
en
OFC: an opportunistic caching system for FaaS platforms
|
Auteur(s) |
Djob Mvondo
1, 2
, Mathieu Bacou
3, 4, 5, 6
, Kevin Nguetchouang
7
, Lucien Ndjie
7
, Stéphane Pouget
2
, Josiane Kouam
8
, Renaud Lachaize
1
, Jinho Hwang
9
, Tim Wood
10
, Daniel Hagimont
11, 12
, Noël de Palma
1
, Bernabé Batchakui
7
, Alain Tchana
2, 7
1
ERODS -
Efficient and Robust Distributed Systems
( 1043311 )
- Laboratoire LIG - Bâtiment IMAG - 700 avenue Centrale, CS 40700 - 38058 Grenoble cedex 9
- France
2
AVALON -
Algorithms and Software Architectures for Distributed and HPC Platforms
( 179717 )
- ENS Lyon 46 allée d'Italie
69364 Lyon Cedex 07
- France
3
TSP - INF -
Département Informatique
( 5170 )
- France
4
IP Paris -
Institut Polytechnique de Paris
( 563936 )
- France
5
ACMES-SAMOVAR -
Architecture, Cloud continuum, formal Models, artificial intElligence and Services in distributed computing
( 442703 )
- TELECOM Sudparis
9 rue Charles Fourier
91011 EVRY
- France
6
Benagil-Samovar -
Efficient and safe distributed systems
( 1179941 )
- France
7
ENSPY -
Ecole Nationale Supérieure Polytechnique de Yaoundé
( 26713 )
- Cameroun
8
TRiBE -
inTeRnet BEyond the usual
( 563449 )
- France
9
Facebook
( 332252 )
- États-Unis
10
GW -
The George Washington University
( 304728 )
- 2121 Eye Street, NW , Washington, DC 20052
- États-Unis
11
IRIT-SEPIA -
Système d’exploitation, systèmes répartis, de l’intergiciel à l’architecture
( 442522 )
- IRIT
118 Route de Narbonne
31062 Toulouse Cedex 9
- France
12
Toulouse INP -
Institut National Polytechnique (Toulouse)
( 448187 )
- France
|
Titre du congrès |
EUROSYS 2021 - 16th European Conference on Computer Systems
|
Date de publication |
2021-04
|
Ville |
Edinburgh (online)
|
Date début congrès |
2021-04-26
|
Date fin congrès |
2021-04-28
|
Source |
EuroSys '21: Proceedings of the Sixteenth European Conference on Computer Systems
|
Actes |
Oui
|
Page/Identifiant |
228-244
|
URL du congrès ou éditeur |
https://dl.acm.org/doi/10.1145/3447786.3456239
|
Langue du document |
Anglais
|
Vulgarisation |
Non
|
Comité de lecture |
Oui
|
Invité |
Non
|
Audience |
Internationale
|
Pays |
Royaume-Uni
|
Collaboration/Projet |
|
Financement |
|
Projet(s) ANR |
|
Éditeur commercial |
|
Domaine(s) |
|
Voir aussi |
|
Mots-clés |
en
Cloud computing, Serverless, Functions as a service (FaaS), Cache, Latency
|
DOI | 10.1145/3447786.3456239 |
Origine :
Fichiers produits par l'(les) auteur(s)
Loading...