TY - JOUR
T1 - Multi-timescale online optimization of network function virtualization for service chaining
AU - Chen, Xiaojing
AU - Ni, Wei
AU - Chen, Tianyi
AU - Collings, Iain B.
AU - Wang, Xin
AU - Liu, Ren Ping
AU - Giannakis, Georgios B.
PY - 2019/12/1
Y1 - 2019/12/1
N2 - Network Function Virtualization (NFV) can cost-efficiently provide network services by running different virtual network functions (VNFs) at different virtual machines (VMs) in a correct order. This can result in strong couplings between the decisions of the VMs on the placement and operations of VNFs. This paper presents a new fully decentralized online approach for optimal placement and operations of VNFs. Building on a new stochastic dual gradient method, our approach decouples the real-time decisions of VMs, asymptotically minimizes the time-average cost of NFV, and stabilizes the backlogs of network services with a cost-backlog tradeoff of [ε, 1/ε], for any ε > 0. Our approach can be relaxed into multiple timescales to have VNFs (re)placed at a larger timescale and hence alleviate service interruptions. While proved to preserve the asymptotic optimality, the larger timescale can slow down the optimal placement of VNFs. A learn-and-adapt strategy is further designed to speed the placement up with an improved tradeoff [ε, log 2(ε)/√ε]. Numerical results show that the proposed method is able to reduce the time-average cost of NFV by 23% and reduce the queue length (or delay) by 74%, as compared to existing benchmarks.
AB - Network Function Virtualization (NFV) can cost-efficiently provide network services by running different virtual network functions (VNFs) at different virtual machines (VMs) in a correct order. This can result in strong couplings between the decisions of the VMs on the placement and operations of VNFs. This paper presents a new fully decentralized online approach for optimal placement and operations of VNFs. Building on a new stochastic dual gradient method, our approach decouples the real-time decisions of VMs, asymptotically minimizes the time-average cost of NFV, and stabilizes the backlogs of network services with a cost-backlog tradeoff of [ε, 1/ε], for any ε > 0. Our approach can be relaxed into multiple timescales to have VNFs (re)placed at a larger timescale and hence alleviate service interruptions. While proved to preserve the asymptotic optimality, the larger timescale can slow down the optimal placement of VNFs. A learn-and-adapt strategy is further designed to speed the placement up with an improved tradeoff [ε, log 2(ε)/√ε]. Numerical results show that the proposed method is able to reduce the time-average cost of NFV by 23% and reduce the queue length (or delay) by 74%, as compared to existing benchmarks.
UR - http://www.scopus.com/inward/record.url?scp=85058143011&partnerID=8YFLogxK
U2 - 10.1109/TMC.2018.2885301
DO - 10.1109/TMC.2018.2885301
M3 - Article
AN - SCOPUS:85058143011
SN - 1536-1233
VL - 18
SP - 2899
EP - 2912
JO - IEEE Transactions on Mobile Computing
JF - IEEE Transactions on Mobile Computing
IS - 12
ER -