An energy saving approach
Résumé
improved in terms of scalability, response efficiency, ease of load balancing
among others. However, it is still a challenge to dynamically
deploy them according to devices’ heterogeneity and energy consumption
concerns, while maintaining a defined QoS. Centralized
and decentralized approaches that manage microservices deployment
have the traditional pros and cons long discussed over time.
While the former offer greater control over distributed application
components, the latter offers frugal network negotiations, no
system-wide crashes, privacy, among others.
This work focuses on identifying “ideal” host candidates for
microservices’ execution in a decentralized network, applying runtime
scheduling operations (migration or duplication) to reduce
energy consumption. To do this, we created a scheduling algorithm
using MAAN (a P2P approach) to interpret a decentralized network
as a multidimensional resource (capacity-demand) space, which
supports range queries in a logarithmic quantity of hops. In this
way, a node that runs a set of microservices is able to 1) map them
in terms of their execution requirements (i.e. CPU frequency, RAM
capacity, Network rate and disk speed) 2) Select an ideal microservice
to be moved or duplicated, 3) find ideal node(s) that meet all
those requirements in an optimal computational complexity and 4)
negotiate the movement or duplication of the selected microservice,
by analyzing energy consumption and QoS criteria.
Mots clés
Computing methodologies → Distributed algorithms
Computer systems organization → distributed algorithms power consumption energy consumption P2P microservices cloud computing
Computer systems organization → distributed algorithms
power consumption
energy consumption
P2P
microservices
cloud computing
Domaines
Informatique [cs]Origine | Fichiers produits par l'(les) auteur(s) |
---|