[Paper Review] PRIMAL: PRofIt Maximization Avatar pLacement for Mobile Edge Computing

Xiang Sun, Nirwan Ansari|arXiv (Cornell University)|Oct 14, 2015

Cloud Computing and Resource Management22 references4 citations

TL;DR

This paper proposes PRIMAL, a profit-maximizing live avatar migration strategy in mobile edge computing that dynamically relocates user avatars across cloudlets to minimize end-to-end (E2E) delay while balancing migration costs. By optimizing the trade-off between E2E delay reduction (gain) and resource overhead from migration (cost), PRIMAL outperforms both 'Follow Me' and 'Static' strategies in maintaining low latency and minimizing migration overhead.

ABSTRACT

We propose a cloudlet network architecture to bring the computing resources from the centralized cloud to the edge. Thus, each User Equipment (UE) can communicate with its Avatar, a software clone located in a cloudlet, with lower end-to-end (E2E) delay. However, UEs are moving over time, and so the low E2E delay may not be maintained if UEs' Avatars stay in their original cloudlets. Thus, live Avatar migration (i.e., migrating a UE's Avatar to a suitable cloudlet based on the UE's location) is enabled to maintain low E2E delay between each UE and its Avatar. On the other hand, the migration itself incurs extra overheads in terms of resources of the Avatar, which compromise the performance of applications running in the Avatar. By considering the gain (i.e., the E2E delay reduction) and the cost (i.e., the migration overheads) of the live Avatar migration, we propose a PRofIt Maximization Avatar pLacement (PRIMAL) strategy for the cloudlet network in order to optimize the tradeoff between the migration gain and the migration cost by selectively migrating the Avatars to their optimal locations. Simulation results demonstrate that as compared to the other two strategies (i.e., Follow Me Avatar and Static), PRIMAL maximizes the profit in terms of maintaining the low average E2E delay between UEs and their Avatars and minimizing the migration cost simultaneously.

Motivation & Objective

• To address the challenge of maintaining low end-to-end (E2E) delay between moving User Equipment (UEs) and their software clones (Avatars) in mobile edge computing.
• To mitigate the performance degradation caused by live Avatar migration, which consumes computational and network resources.
• To develop a strategy that optimally balances the gain (reduced E2E delay) and cost (migration overhead) of Avatar migration.
• To evaluate and compare PRIMAL against two baseline strategies: 'Follow Me' (prioritizing low delay) and 'Static' (prioritizing low migration cost).

Proposed method

• PRIMAL formulates the Avatar placement problem as a profit maximization optimization, where profit is defined as the difference between migration gain (E2E delay reduction) and migration cost (resource overhead).
• The profit function incorporates a tunable weight parameter α to balance the trade-off between delay reduction and migration cost, allowing dynamic adjustment based on network conditions.
• The algorithm uses a predictive model to estimate future UE locations and selects optimal cloudlet migration targets to minimize E2E delay while avoiding unnecessary migrations.
• Migration decisions are made based on real-time monitoring of UE mobility, cloudlet resource utilization, and E2E delay measurements across the network.
• The strategy leverages a Software-Defined Networking (SDN)-based core network to enable flexible and fast path provisioning between Avatars and UEs.
• Simulations are conducted using realistic mobility traces to evaluate performance across varying values of α, measuring average RTT, number of migrations, and migration time.

Experimental results

Research questions

• RQ1 How can Avatar migration be strategically managed to maintain low E2E delay for moving UEs in a cloudlet network?
• RQ2 What is the optimal trade-off between E2E delay reduction (gain) and resource consumption from migration (cost) in mobile edge computing?
• RQ3 How does the proposed PRIMAL strategy compare to static and 'Follow Me' migration strategies in terms of profit, delay, and migration overhead?
• RQ4 How does the weight parameter α affect the balance between delay reduction and migration cost in PRIMAL?

Key findings

• PRIMAL achieves the highest profit by effectively balancing E2E delay reduction and migration cost, outperforming both the 'Follow Me' and 'Static' strategies.
• As α increases, the average profit gap between PRIMAL and Static decreases, indicating that PRIMAL converges toward the Static strategy when migration cost dominates.
• When α is low, PRIMAL triggers more migrations to reduce RTT, even at the cost of higher migration time, demonstrating responsiveness to delay sensitivity.
• As α increases, PRIMAL reduces unnecessary migrations that offer marginal delay improvement but high migration cost, showing adaptive cost control.
• The simulation results confirm that PRIMAL achieves the best trade-off between average RTT and number of migrations, with a clear trend in performance improvement over baselines.
• The proposed strategy effectively reduces the volume of data transmitted to remote data centers by enabling local data processing at cloudlets, enhancing scalability and reducing latency.