Ali Jawad
FAHS

Cloud computing infrastructures are very powerful and flexible, but they are also located very far from their end users. Typical network latencies between an end user and the closest public cloud data center are in the order of 20-40 ms over high-quality wired networks, and 100-150 ms over 4G mobile phone connections. This performance level is acceptable for simple applications such as web browsing, but it makes it impossible to create a wide range of interactive applications. For example, to enable an "instantaneous" feeling, augmented reality applications require that end-to-end latencies (including all networking and processing delays) remain below 20 ms.

To address these issues, a new type of "fog computing" infrastructures is being designed [1,5]. Instead of treating the mobile operator's network as a high-latency dumb pipe between the end users and the external service providers, fog platforms aim to bring cloud resources at the edge of the network, in very close physical proximity with the end users. This is expected to offer extremely low latency between the client devices and the cloud resources serving them.

The goal of this project is to reduce the discrepancy between the broadly distributed compute/storage resources and the -- currently -- extremely centralized control of these resources. We can exploit the fact that the virtual resources in a fog computing platform are in most cases created in immediate proximity of the user(s) who will access them. In this perspective, the platform management processes could be distributed evenly across the infrastructure nodes so that the virtual and physical resources, the users accessing them, and the management processes organizing this system, will be co-located within a few hundred meters from each other. One interesting direction to address this problem -- which remains to be (in)validated by the doctoral student -- is to execute cloud resource scheduling algorithms [7] on every point-of-presence of the system (whereas traditional clouds centralize these algorithms in a single node), and to base the necessary coordination of multiple schedulers on gossiping algorithms [6] between neighboring points-of-presence.

This project is being conducted within the IRISA Myriads team which is working on the design of innovative infrastructures and middleware for future fog computing platforms [2,3,4].

References

• [1] "MEC-ConPaaS: An experimental single-board based mobile edge cloud." Alexandre van Kempen, Teodor Crivat, Benjamin Trubert, Debaditya Roy and Guillaume Pierre. In Proceedings of the IEEE Mobile Cloud conference, April 2017.
• [2] "Kangaroo: A Tenant-Centric Software-Defined Cloud Infrastructure." Kaveh Razavi, Ana Ion, Genc Tato, Kyuho Jeong, Renato Figueiredo, Guillaume Pierre and Thilo Kielmann. In Proceedings of the IEEE International Conference on Cloud Engineering (IC2E), Tempe, AZ, USA, March 2015. http://www.globule.org/publi/KTCSDCI_ic2e2015.html
• [3] "ConPaaS: a Platform for Hosting Elastic Cloud Applications." Guillaume Pierre and Corina Stratan. IEEE Internet Computing 16(5), September-October 2012. http://www.globule.org/publi/CPHECA_ic2012.html
• [4] "The mobile edge cloud testbed at IRISA Myriads team." https://youtu.be/7uLkLitiSPo
  
• [5] "Fog Computing and its Ecosystem." Ramin Elahi, tutorial at the USENIX FAST conference, 2016. http://bit.do/c2Ta7<
• [6] "Gossip-based peer sampling." Mark Jelasity, Spyros Voulgaris, Rachid Guerraoui, Anne-Marie Kermarrec and Maarten Van Steen. ACM Transactions on Computer Systems 25(3), 2007.
  http://acropolis.cs.vu.nl/~spyros/www/papers/Gossi...
• [7] "A Survey on Resource Scheduling in Cloud Computing: Issues and Challenges." Sukhpal Singh and Inderveer Chana. Journal Grid Computing 14, 2016. http://link.springer.com/article/10.1007/s10723-01...

Abstract:
Latency-sensitive fog computing applications may use replication both to scale their capacity and to place application instances as close as possible to their end users. In such geo-distributed environments, a good replica placement should maintain the tail network latency between end-user devices and their closest replica within acceptable bounds while avoiding overloaded replicas. When facing non-stationary workloads it is essential to dynamically adjust the number and locations of a fog application's replicas. We propose Voilà, a tail-Iatency-aware auto-scaler integrated in the Kubernetes orchestration system. Voila maintains a fine-grained view of the volumes of traffic generated from different user locations, and uses simple yet highly-effective procedures to maintain suitable application resources in terms of size and location.

Abstract:
Latency-sensitive applications often use fog computing platforms to place replicas of their services as close as possible to their end users. A good placement should guarantee a low tail network latency between end-user devices and their closest replica while keeping the replicas load balanced. We propose a latency-aware scheduler integrated in Kubernetes which uses simple yet highly-effective heuristics to identify suitable replica placements, and to dynamically update these placements upon any evolution of user-generated traffic.

Abstract:
Container orchestration engines such as Kubernetes do not take into account the geographical location of application replicas when deciding which replica should handle which request. This makes them ill-suited to act as a general-purpose fog computing platforms where the proximity between end users and the replica serving them is essential. We present proxy-mity, a proximity-aware traffic routing system for distributed fog computing platforms. It seamlessly integrates in Kubernetes, and provides very simple control mechanisms to allow system administrators to address the necessary trade-off between reducing the user-to-replica latencies and balancing the load equally across replicas. proxy-mity is very lightweight and it can reduce average user-to-replica latencies by as much as 90% while allowing the system administrators to control the level of load imbalance in their system.

Résumé :
Les réseaux multi-sauts 802.15.4e TSCH s'appuient sur des échéanciers de communication efficaces. Notre solution complète les algorithmes de construction de tels échéanciers en limitant la réutilisation de cellules temps-fréquence déjà utilisées par des noeuds voisins. Nos simulations montrent un gain significatif par rapport à l'existant.

Abstract:
The IEEE802.15.4e standard for low power wireless sensor networks defines a new mode called Time Slotted Channel Hopping (TSCH) as Medium Access Control (MAC). TSCH allows highly efficient deterministic time-frequency schedules that are built and maintained by the 6TiSCH operation sublayer (6top). In this paper, we propose a solution to limit the allocation of identical cells to co-located pair of nodes by distributed TSCH scheduling algorithms. It consists of making nodes able to overhear past cell negotiations exchanged in shared cells by their neighbors and prevent the nodes from reusing already assigned cells in future allocations. Our mechanism has been tested through simulations that show a significant improvement with respect to random scheduling algorithms.