Leonardo Marques Epifânio
Cathode: A Consistency-Aware Data Placement Algorithm for the Edge
Tese submetida para provas de mestrado em Engenharia
Informática e de Computadores Instituto Superior
Técnico, Universidade de Lisboa.
Abstract
Placing data replicas in edge nodes is a key strategy to offer low
latency to clients and to improve network utilization. In the case of
immutable objects, data placement is only constrained by the limited
capacity of edge nodes. However, for mutable objects, one also needs
to consider the cost of keeping replicas consistent, which then varies
with the data consistency model: a replica placement algorithm that
performs well for weakly consistent replicas may perform poorly when
strong consistency is required. In this thesis, we present Cathode, a
replica placement algorithm that is tailored for the requirements of
the edge environment, all the while being consistency-aware, making
placement decisions based on client demand, storage costs, and the
costs of keeping replicas consistent. In the underlying system,
different objects may use different consistency models and Cathode
makes placement decisions accordingly. Because optimal replica
placement is known to be an NP-hard problem, Cathode resorts to an
heuristic that is decentralized and scalable, providing fast
convergence, but also achieving high quality deployments. The
extensive performance evaluation reported in this thesis shows that it
outperforms previous state-of-the-art replica placement algorithms.
Publicações
- Cathode: A Consistency-Aware Data Placement Algorithm
for the Edge
- Leonardo Marques
Epifânio
- MSc Thesis. Instituto Superior Técnico,
Universidade de Lisboa.
- January 2021.
- Available BibTeX, MSC Thesis, and extended abstract, and
mid-term
report.
- Cathode: A Consistency-Aware Data
Placement Algorithm for the Edge.
- L. Epifânio, C. Correia
and L. Rodrigues.
- Proceedings of the 20th IEEE International
Symposium on Network Computing and Applications (NCA 2021), Online,
November, 2021.
Luís Rodrigues