MobiEdgeSim: A Simulator for Large-Scale Mobile MEC Server

Scenarios

Tianhao Zhang

, Owen Gallagher

, Aqeel H. Kazmi

and Siobh

an Clarke

School of Computer Science and Statistics, Trinity College Dublin, College Green, Dublin 2, Ireland

Keywords:

Multi-Access Edge Computing, Service Placement, Dynamic Scheduling, Resource Allocation.

Abstract:

Multi-access edge computing (MEC) is an emerging network architecture that brings computational resources

closer to users, enabling localized computation and real-time task responses. While numerous simulators have

been developed to explore MEC environments, most assume static MEC servers and focus on user mobility.

However, this static assumption limits the exploration of mobile MEC servers and their potential beneﬁts in

dynamic environments. In this paper, we present MobiEdgeSim, a simulation framework for large-scale static

and mobile MEC server scenarios, where mobile MEC servers may be deployed on buses, trams, trains or

other mobile vehicles. The simulator is built on top of the OMNeT++ and Simu5G frameworks, integrating

SUMO for realistic road trafﬁc simulations and Veins for seamless mobility and communication modelling.

The framework supports large-scale simulations, conﬁgurable scenarios, complex network design, dynamic

mobile simulations based on real-world transportation systems, and evaluation of matrices under diverse con-

ditions. By introducing mobility-aware MEC server designs, this work enables researchers to study complex

urban environments, and optimize resource efﬁciency in large-scale mobile networks. The performance of

MobiEdgeSim is evaluated under varying scenarios and service placement strategies.

1 INTRODUCTION

The advent of ﬁfth-generation (5G) networks has

brought revolutionary improvements in communica-

tion technology, characterized by ultra-low latency,

high bandwidth, and real-time access to radio net-

work information. These features have unlocked new

opportunities for deploying applications and services

closer to end users by leveraging the Radio Access

Network (RAN) edges. This paradigm shift, known

as Multi-access Edge Computing (MEC), enhances

computational efﬁciency and responsiveness by re-

ducing the dependency on centralized data centres.

However, the dynamic nature of 5G networks, with

highly mobile user equipment (UE) and even mobile

MEC servers, introduces signiﬁcant challenges. As

illustrated in Figure 1, the mobility of both users and

servers increases the complexity of resource manage-

ment tasks, such as service placement and allocation,

requiring advanced computational strategies to ensure

optimal system performance in real-time.

Service placement plays a critical role in ensur-

ing Quality of Service (QoS) in MEC environments.

https://orcid.org/0009-0004-0932-0278

https://orcid.org/0009-0008-9795-0390

https://orcid.org/0000-0002-8365-9892

https://orcid.org/0000-0001-5721-9976

Orchestrator

Edge

Node

Edge Layer

Edge

Node

Mobile

Edge

Node

Mobile

Edge

Node

Devices &

Users

Scheduler

Broker &

Compute

Figure 1: System overview: dynamic service placement in

static and mobile MEC nodes.

Researchers and MEC network managers must thor-

oughly evaluate various placement strategies under

realistic conditions to optimize performance, neces-

sitating robust simulation tools. Applications and ser-

vices are deployed to various edge locations based

on real-time network conditions, geographical loca-

tions, service types, and resource availability. These

dynamic factors make it imperative to develop efﬁ-

cient strategies for task distribution to optimize per-

formance while maintaining QoS under varying con-

ditions.

Simulators are indispensable tools for researchers

navigating the complexities and dynamic nature of

5G-based services. Given the unpredictable nature of

network conditions, user mobility, and uneven service

Zhang, T., Gallagher, O., Kazmi, A. H. and Clarke, S.

MobiEdgeSim: A Simulator for Large-Scale Mobile MEC Server Scenarios.

DOI: 10.5220/0013529300003970

In Proceedings of the 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications (SIMULTECH 2025), pages 81-92

ISBN: 978-989-758-759-7; ISSN: 2184-2841

distribution, testing in real-world environments is of-

ten challenging. By utilizing simulators, researchers

can effectively model and evaluate resource man-

agement techniques, analyse the feasibility of novel

algorithms, and assess the performance of latency-

sensitive applications under diverse scenarios.

Simulators play a critical role in studying MEC

environments, yet existing tools face signiﬁcant lim-

itations in modelling dynamic and complex scenar-

ios. For instance, INET (M

esz

aros et al., 2019),

an OMNeT++-based framework, excels in network-

level simulations but lacks integration with mobility

models, limiting its applicability to dynamic user or

server scenarios. Similarly, NS-3 (Riley and Hen-

derson, 2010), widely used for protocol-level studies,

cannot incorporate real-world road trafﬁc conditions,

which are essential for capturing the impact of con-

gestion or smooth trafﬁc ﬂows on MEC performance.

Veins (Sommer et al., 2019) and MAACO (Cabrera

et al., 2022) provide partial solutions by integrating

mobility with network communication but remain fo-

cused on user mobility, overlooking the potential of

server mobility, such as deploying MEC servers on

buses or trains.

Figure 2: Mobile MEC server scenarios.

These limitations are particularly pronounced in

large-scale urban networks, where dynamic interac-

tions between mobile users, real-time trafﬁc condi-

tions, and computational resources play a crucial role.

As illustrated in Figure 2, smart city applications

such as AR support for people with disabilities, real-

time navigation services, and autonomous driving re-

quire simulators that support scalability and ﬂexibility

while incorporating realistic mobility and communi-

cation dynamics. However, most existing tools can-

not model these scenarios effectively, as they assume

static MEC servers and lack support for advanced ap-

plication designs across diverse environments.

To address these gaps, we propose a novel simu-

lation framework called MobiEdgeSim, whose source

code is publicly available on GitHub (Zhang et al.,

2025), which overcomes these limitations by integrat-

ing static and mobile MEC servers, realistic road traf-

ﬁc dynamics, and large-scale simulation capabilities,

and provides ﬂexibility for evaluating diverse applica-

tion scenarios. Built on top of OMNeT++, Simu5G,

Veins, and SUMO, our framework supports investiga-

tions into mobility-aware service placement, resource

allocation, and system performance under highly dy-

namic conditions. Through detailed conﬁgurations of

urban and vehicular mobility, we allow researchers

to explore how servers in motion can efﬁciently re-

spond to shifting network demands and reduce re-

sponse times for critical applications.

The rest of this paper is organized as follows. Sec-

tion 2 reviews related work and highlights the gaps

in existing MEC simulation frameworks. Section 3

presents our simulation framework design, elaborat-

ing how we integrate Simu5G, Veins, and SUMO un-

der OMNeT++. Section 4 describes the modelling

approach for both static and mobile nodes, including

MEC server extensions and resource allocation mech-

anisms. Section 5 evaluates the performance of the

framework under varying scenarios and service place-

ment strategies, demonstrating the beneﬁts of intro-

ducing mobile MEC servers. Finally, Section 6 con-

cludes the paper with a summary of key ﬁndings and

outlines future research directions.

2 RELATED WORK

We conducted a thorough analysis of both cloud

and network simulators to identify effective solutions

for modelling large-scale mobile Edge Computing

(MEC) server scenarios.

In the category of cloud simulators, CloudSim is

a widely used simulation toolkit designed for mod-

elling and evaluating resource provisioning strategies

in centralized cloud computing environments (Cal-

heiros et al., 2011). Its modular design and exten-

sibility make it a popular choice among researchers.

CloudSim was initially developed for static cloud data

centers, making it inherently unsuitable for edge com-

puting and Internet of Things (IoT) scenarios. This

limitation means it does not natively support features

like mobile user equipment (UE), distributed edge

nodes, or dynamic resource allocation.

Several extensions to CloudSim aim to incorpo-

rate features more relevant to edge and fog comput-

ing: iFogSim (Gupta et al., 2017) adds a fog layer on

top of CloudSim, enabling IoT device modelling and

task scheduling, but lacks native support for complex

mobility of edge nodes and realistic network-layer in-

teractions. EdgeCloudSim (Sonmez et al., 2018) ex-

tends CloudSim with mobility models and simpliﬁed

communication protocols to evaluate edge resource

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

ofﬂoading. However, it focuses on static edge infras-

tructure and does not address mobile MEC servers.

IoTSim-Edge (Nandan Jha et al., 2019) includes fea-

tures for energy modelling, heterogeneous communi-

cation protocols, and directed acyclic graph (DAG)

application structures. Despite improved support for

IoT-edge interactions, it does not natively handle the

movement of MEC nodes or integrate trafﬁc condi-

tions at scale.

YAFS (Lera et al., 2019) (Yet Another Fog Sim-

ulator) is a Python-based framework emphasizing re-

source management in fog and edge infrastructures.

Its customizable topologies and dynamic application

deployments facilitate latency, energy, and network

utilization analysis. However, YAFS is not designed

for protocol-level experiments or integration with re-

alistic vehicular or road trafﬁc simulators.

FogNetSim++ (Qayyum et al., 2018), built on top

of OMNeT++, supports mobility models, resource

scheduling, and handover mechanisms in fog envi-

ronments. Nevertheless, it does not directly incor-

porate advanced road trafﬁc simulation tools such as

SUMO, limiting its applicability for urban-scale or

highly mobile edge scenarios. Moreover, FogNet-

Sim++ remains focused on user mobility rather than

the mobility of MEC servers themselves.

In the category of network simulators, NS-3 (Ri-

ley and Henderson, 2010) and OMNeT++ (Varga

and Hornig, 2010) are both prominent discrete-event

network simulators, extensively used for protocol-

level research in wired, wireless, and vehicular net-

works. They provide low-level control over net-

working stacks and support a range of mobility mod-

els, making them ideal for studying communication

performance. However, their core design does not

directly address higher-level aspects such as task

scheduling, dynamic service placement, or orches-

tration of edge resources. Add-on frameworks like

Veins (Sommer et al., 2019) (for vehicular networks

in OMNeT++) enhance mobility and network simula-

tion capabilities. While these extensions support user

or vehicle mobility, they do not fully capture resource

management or dynamic computations at the edge,

particularly when MEC servers themselves are in mo-

tion. Simu5G is another extension of OMNeT++ de-

signed to simulate 5G New Radio (NR) networks. It

integrates with the ETSI MEC architecture to sup-

port the evaluation of MEC service placement and re-

source allocation strategies in standardized 5G envi-

ronments.

In summary, although existing simulation tools

and frameworks offer robust capabilities for either

network-layer evaluations or static resource manage-

ment, they generally lack integrated support for large-

scale scenarios where both users and MEC servers are

mobile. Such environments, especially in 5G and be-

yond networks, require a holistic approach that fuses

realistic trafﬁc mobility with dynamic service place-

ment and resource orchestration. We propose Mo-

biEdgeSim, a new simulation framework that builds

on Simu5G, Veins, and SUMO. This framework mod-

els static and mobile MEC servers at scale, allowing

for large-scale urban simulations that effectively ac-

count for mobile server nodes and realistic trafﬁc con-

ditions.

3 SIMULATOR FRAMEWORK

DESIGN

OMNeT++

Veins

Mobility

Road Traffic Simulation

SUMO

INET

Simu5G

Network Development

Network Protocols

Veins Manager

MobiEdgeSim

5G network simulation

Mobile User

Mobile Nodes

Service Placement

Location Control

Network Description

Graphical Interface

Simulation Control

Statistical Analysis

Application Design

MEC extensions

Communication Features

Figure 3: An overview of MobiEdgeSim architecture.

Figure 3 provides an overview of MobiEdgeSim,

which uniﬁes Simu5G, Veins, and SUMO within the

OMNeT++ environment. The objective is to provide

an integrated solution that captures both network-

ing protocols and real-world trafﬁc dynamics, mak-

ing it well-suited for large-scale MEC scenarios that

involve both static and mobile MEC servers. The

framework integrates mechanisms for the rapid re-

location of MEC nodes, facilitating experiments tai-

lored to latency-sensitive applications where the goal

is to effectively alleviate network congestion and re-

duce overall latency. In addition, it uses intricate vehi-

cle mobility patterns to accurately simulate city-wide

trafﬁc ﬂuctuations, demonstrating the impact of these

variations on communication throughput and resource

utilization. The framework also serves as an essential

resource for researchers, allowing them to conduct

simulations related to adaptive resource scheduling

and service placement problems. It enables testing

of solutions aimed at maintaining QoS requirements

in the event of sudden changes in network topology.

To accomplish the objective, the environment

combines a comprehensive 5G NR model (Simu5G),

realistic vehicular traces (SUMO), and a unifying

trafﬁc-network simulation module (Veins). By ex-

MobiEdgeSim: A Simulator for Large-Scale Mobile MEC Server Scenarios

tending this stack to support mobile MEC nodes,

the framework enables in-depth investigations into

mobility-aware service placement, resource manage-

ment, and orchestration strategies in urban and high-

way settings. This holistic approach ensures that

varying user demands, shifting network topologies,

and realistic trafﬁc conditions are all accounted for

when evaluating MEC related scenarios at large

scales.

3.1 Core Simulation Tools

Simu5G is an OMNeT++ based framework that im-

plements a detailed 5G New Radio (NR) model, in-

cluding physical layer abstractions, protocol stacks,

and radio resource management (Nardini et al., 2020).

It allows simulation of ultra-low latency and high-

bandwidth cellular communications, which are fun-

damental features in modern MEC deployments.

Simu5G provides a detailed 5G New Radio (NR)

model, including realistic scheduling, handover pro-

cedures, and channel propagation effects. Moreover,

it incorporates basic modeling capabilities for Multi-

access Edge Computing (MEC), following the ETSI

standards, allowing simulations of static MEC host

deployment and service placement strategies (Noferi

et al., 2023). This level of ﬁdelity makes Simu5G the

backbone for evaluating how diverse service place-

ment or ofﬂoading strategies inﬂuence end-to-end

performance, such as latency, throughput, and packet

delivery ratio, in complex 5G scenarios.

Veins is a popular vehicular network simulation

framework that runs on top of OMNeT++, focusing

on mobility modelling for vehicles, V2X (Vehicle to

Everything) communication protocols, and vehicular

ad hoc networks (Sommer et al., 2019). By connect-

ing to a road trafﬁc simulator like SUMO (Simulation

of Urban Mobility), Veins can inject realistic vehicle

trajectories, speed proﬁles, and trafﬁc conditions into

OMNeT++. This synergy enables the simulation of

large-scale vehicular scenarios, from small city grids

to highway systems.

SUMO is a microscopic, open-source road traf-

ﬁc simulator that provides detailed mobility data for

each individual vehicle, including acceleration, de-

celeration, lane changes, and interactions with trafﬁc

lights (Lopez et al., 2018). Integrating SUMO with

OMNeT++ (via Veins) allows a highly accurate rep-

resentation of on-road mobility patterns, congestion

hot spots, and signal timing effects.

Together, Simu5G, Veins, and SUMO offer a

robust foundation for simulating both the network-

ing and mobility aspects essential for MEC research.

However, to address the unique challenges posed by

mobile MEC servers, additional extensions and mod-

ules were developed.

3.2 Framework Extensions

In order to enhance the capabilities of Simu5G, Veins,

and SUMO for mobile MEC server scenarios, we

developed several extensions in Simu5G. Figure 4

summarizes the key contributions of these extensions,

while the following subsections discuss each in detail.

Figure 4: Key framework extensions developed to enhance

the capabilities of Simu5G.

3.2.1 Mobile MEC Servers

To enable mobility for MEC servers, we introduced

a Cellular module within the UPF (User Plane Func-

tion) submodule of the MEC server. This Cellular

module is designed to receive wireless signals and

establish essential communication links required for

MEC operations, enabling mobile MEC servers to

seamlessly connect with the user equipment and ex-

change radio network information.

Furthermore, we integrated Veins’ mobility mod-

ule into the MEC server component, allowing MEC

nodes to move dynamically across simulated road net-

works. This integration utilizes Veins’ mobility inter-

faces to emulate realistic movement patterns, such as

vehicles, trams, or trains, ensuring that the simulation

closely reﬂects real-world urban mobility scenarios.

In addition, the routing table configurator mod-

ule was updated to an auto-configurator to ac-

commodate the unique initialization requirements of

mobile MEC nodes in OMNeT++. Unlike traditional

conﬁgurators in INET, which are designed primarily

for static nodes, the auto-conﬁgurator is capable of

dynamically initializing routing tables for nodes that

appear after the initial setup phase. This feature is es-

sential for mobile nodes generated through Veins, as

their initialization process differs from that of static

nodes. By leveraging the auto-conﬁgurator, the rout-

ing tables are correctly conﬁgured for mobile MEC

nodes, ensuring seamless integration and reliable op-

eration within the simulation environment.

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

3.2.2 Static MEC Server Enhancements

For static MEC server modules, we incorporated lat-

itude and longitude parameters and mapped them to

speciﬁc locations on the SUMO-generated road net-

work during the creation of the NED ﬁle. By as-

signing these geographical coordinates, each static

MEC server is precisely positioned in the simula-

tion environment corresponding to its designated lo-

cation on the SUMO map. This setup ensures accu-

rate alignment between the simulated network topol-

ogy and real-world geographical layouts, enabling re-

alistic service placement scenarios.

When researchers design their own networks, this

approach provides ﬂexibility to deﬁne static MEC

server locations based on their speciﬁc SUMO road

networks and architecture requirements. Researchers

can strategically place static MEC servers at key

points, such as near intersections, public transport

hubs, or high-demand areas, to simulate various ur-

ban or rural network designs. These predeﬁned coor-

dinates are passed to the MEC orchestrator, which can

allow investigations into making informed decisions

about service placement, resource allocation, and task

scheduling.

This enhancement, along with the Veins mobil-

ity module integrated into mobile MEC servers as

mentioned in the last section, ensures that mobility-

aware algorithms can efﬁciently utilize both static

and mobile nodes. By leveraging the precise loca-

tion data, the orchestrator is equipped with additional

decision-making metrics, such as geographical prox-

imity and spatial distribution of static MEC servers.

This enhancement allows the orchestrator to make

more informed and context-aware decisions, allowing

researchers to test and improve the efﬁciency and per-

formance of mobility-aware algorithms in the simula-

tion.

3.2.3 Orchestrator Data Collection and Node

Selection

We extended the MEC Orchestrator’s capabilities

by implementing more comprehensive data-gathering

mechanisms, enabling more advanced node selection

decisions. Instead of relying solely on simple metrics

such as available CPU or RAM, the Orchestrator now

collects and aggregates a broader set of parameters

from both the network and application layers. Specif-

ically:

• Latency between User and MEC Servers. At

ﬁxed intervals, the system measures the one-way

latency between user equipment (UE) and their

serving MEC nodes by transmitting timestamped

UDP packets. The difference between the em-

bedded timestamp and the arrival time is com-

puted, ensuring the Orchestrator maintains up-to-

date delay information under dynamic mobility

and load conditions.

• Geographical Distance. Both mobile MEC

servers and UEs track their latitude and longitude

through Veins and get static MEC servers position

through the parameter presetting. This enables the

Orchestrator to accurately calculate distances for

location-aware scheduling heuristics, facilitating

the decisions.

• User-Required Resources. Detailed CPU, RAM,

and disk requirements for each incoming ser-

vice request prevent tasks from being assigned to

under-provisioned nodes.

• MEC Servers’ Current Resources. Real-time

snapshots of available RAM, disk space, and CPU

cycles across all MEC servers (both static and mo-

bile) would allow the Orchestrator to balance load

more effectively and mitigate overload scenarios.

By aggregating and analysing these richer data

sources, the enhanced Orchestrator can adapt to sud-

den changes in network load, user distribution, or

server mobility. Moreover, this expanded data col-

lection opens the door for testing more sophisticated

algorithms (e.g., multi-objective optimization or ma-

chine learning) that demand a global, ﬁne-grained

view of the network state. Consequently, the up-

graded Orchestrator provides a ﬂexible foundation for

researchers exploring diverse scheduling and resource

allocation strategies under realistic urban-scale sce-

narios.

3.3 Framework Integration

Our simulation framework leverages SUMO for road

trafﬁc simulation and OMNeT++ as the discrete event

simulation platform. Veins, as a vehicular net-

work simulation framework, bridges the gap between

SUMO and OMNeT++, enabling real-time mobil-

ity updates through TraCI (Trafﬁc Control Interface).

Meanwhile, INET provides fundamental networking

models, including IP-based communication and wire-

less protocols, which are required for Simu5G’s oper-

ation within OMNeT++.

Simu5G, which models 5G New Radio (NR) com-

munication, operates within OMNeT++ using INET’s

network stack and interacts with Veins’ mobility mod-

els to ensure accurate representation of MEC server

movement. As MEC nodes are deployed on vehi-

cles, Simu5G dynamically updates their connectiv-

ity status, ensuring seamless handovers between 5G

MobiEdgeSim: A Simulator for Large-Scale Mobile MEC Server Scenarios

base stations and allowing latency-sensitive applica-

tions to operate under mobile conditions. Our ex-

tended Simu5G framework builds upon INET and

Veins INET (a subproject of Veins) to utilize realistic

mobility data from SUMO through Veins via TraCI.

Although this integration leverages existing tools, our

primary contribution is the introduction of mobile

MEC servers to Simu5G. Speciﬁcally, by enabling

MEC servers to dynamically follow the mobility pat-

terns provided by SUMO, our extensions signiﬁcantly

enrich the types of scenarios researchers can simulate.

The successful integration of these components

allows researchers to simulate real-world MEC de-

ployment scenarios with a high degree of realism.

Through this framework, users can evaluate various

task placement strategies, resource allocation poli-

cies, and network conﬁgurations under diverse ur-

ban mobility conditions. Building on these exten-

sions, our simulation framework seamlessly inte-

grates Simu5G, Veins, and SUMO with newly devel-

oped modules for mobile MEC servers. This inte-

gration process involves deﬁning urban environments

within SUMO to generate realistic trafﬁc patterns and

vehicle movements, utilizing SUMO’s detailed mo-

bility data to drive Veins’ vehicular models for an ac-

curate representation of vehicle behaviours and inter-

actions, and deploying MEC servers on selected vehi-

cles, trams, or trains, each following mobility patterns

governed by Veins’ mobility module. In the next sec-

tion, we provide a detailed description of how a re-

searcher can design their simulation and integrate our

simulation framework to support their required capa-

bilities.

4 MODELLING A SIMULATION

This section outlines the workﬂow for modelling and

executing a simulation scenario using MobiEdgeSim.

Figure5 illustrates the key steps in setting up and

running a large-scale simulation scenario in MobiEd-

geSim, from selecting and processing the raw Open-

StreetMap (OSM) data to conﬁguring the OMNeT++

environment and collecting performance metrics. We

will elaborate on each step in detail in the follow-

ing sections, providing researchers with the necessary

guidance to replicate or customize large-scale sim-

ulations that integrate both static and mobile MEC

servers, enabling comprehensive evaluations.

4.1 Road Trafﬁc Modelling with SUMO

Realistic road trafﬁc modelling is essential for accu-

rately assessing the behaviour of mobile nodes, in-

Figure 5: Workﬂow of simulation modelling.

cluding vehicles equipped with MEC servers. To

achieve this, SUMO (Simulation of Urban Mobility)

provides a high-ﬁdelity, microscopic trafﬁc simula-

tion framework capable of generating individual ve-

hicle movements based on real-world road networks.

However, to create realistic urban trafﬁc scenarios,

SUMO requires detailed road network data as input.

OpenStreetMap (OSM) is an open-access geo-

graphic database that provides comprehensive road

network information and is a valuable data source for

SUMO-based simulations (OpenStreetMap contribu-

tors, 2017). However, OSM data is not directly com-

patible with SUMO and must ﬁrst be converted into

SUMO’s .net.xml format. This conversion can be

performed using netconvert, a SUMO tool designed

to process OSM ﬁles and generate road networks that

adhere to SUMO’s simulation framework.

To further simplify this process, SUMO provides

an automated tool, osmWebWizard.py, which directly

retrieves OSM data, processes it with netconvert,

and generates a complete simulation scenario. This

tool allows users to select a geographic region of in-

terest, specify vehicle types, and automatically cre-

ate all necessary SUMO ﬁles, including the road net-

work (.net.xml), vehicle routes (.rou.xml), and a

road simulation conﬁguration ﬁle (.sumocfg). By

eliminating the need for manual data downloading

and conversion, osmWebWizard.py signiﬁcantly re-

duces the complexity of setting up SUMO-based sim-

ulations, making it particularly useful for large-scale

trafﬁc studies and urban mobility analysis.

For users requiring more control over network

conversion parameters, netconvert can be used in-

dependently. This allows for ﬁne-tuned adjustments

such as ﬁltering speciﬁc road types, modifying in-

tersection behaviours, or inferring trafﬁc light place-

ments using additional parameters. Both approaches

ensure the accurate integration of real-world road

infrastructure into SUMO, enabling researchers to

model realistic trafﬁc conditions while maintaining

ﬂexibility in scenario design.

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

After generating the SUMO scenario, Veins estab-

lishes a TraCI connection that continuously retrieves

mobility updates such as vehicle positions, speeds,

and lane changes, and relays them to OMNeT++ in

real time. This approach allows both user equip-

ment and mobile MEC servers to move along realistic

road trajectories that reﬂect actual trafﬁc conditions

in the network simulation. Meanwhile, Simu5G in-

tegrates 5G New Radio (NR) communication mod-

els into OMNeT++, ensuring that both the mobil-

ity and the details of the network layer are cap-

tured. Once the modules are properly conﬁgured, re-

searchers can ﬁne-tune additional parameters, includ-

ingg node placement, resource constraints, and traf-

ﬁc densities, before proceeding to environment con-

ﬁguration and performance evaluation. Section 4.2

describes how to modify these parameters using the

NED (Network Description) ﬁles and the INI (Ini-

tialization) ﬁles and outlines the general procedure

for executing experiments and collecting performance

metrics.

4.2 Environment Conﬁguration

The simulation environment conﬁguration is handled

primarily through two key ﬁles: the NED ﬁle and

the INI ﬁle, which allow researchers to design net-

work topologies, specify simulation parameters, and

deﬁne node behaviour according to their experimen-

tal requirements.

The NED ﬁle is used to construct the network

topology by specifying various simulation compo-

nents. Users can deﬁne different numbers of User

Equipment (UE), gNB base stations, UPF (User Plane

Function), and MEC servers. Static nodes, such as

gNBs, UPFs, and static MEC servers, are explicitly

placed in the network topology and do not change

position during the simulation. In contrast, dynamic

nodes, including UE and mobile MEC hosts, are de-

ﬁned as vector objects and their movement is gov-

erned by SUMO and Veins. This ensures that the

number, location, and mobility of these nodes are

dynamically controlled according to a SUMO trafﬁc

simulation.

The INI ﬁle is used to conﬁgure simulation run-

time settings, application parameters, Veins-speciﬁc

conﬁgurations, and other module-speciﬁc parameters.

It allows researchers to customize by adjusting the to-

tal execution time, deﬁning mobility behaviours, set-

ting resource constraints for MEC servers, and spec-

ifying communication parameters for different net-

work elements. The INI ﬁle also ensures proper

integration with SUMO by deﬁning the connection

through Veins’ parameters, enabling seamless mobil-

ity updates. By modifying the INI ﬁle, users can

ﬁne-tune their experiments to reﬂect different net-

work conditions and application requirements without

altering the underlying network model.

By modifying these conﬁguration ﬁles, users can

design and execute experiments tailored to their spe-

ciﬁc research objectives. This ﬂexibility enables sim-

ulations of varying complexity, enabling researchers

to experiment with both small-scale test beds and

large-scale urban deployments.

4.3 Execution and Data Collection

OMNeT++ offers a built-in statistical framework for

collecting simulation results and logging runtime pa-

rameters. The simulator supports various output for-

mats, including scalar and vector ﬁles, which can be

automatically generated based on predeﬁned collec-

tion rules speciﬁed in the INI ﬁle. These ﬁles store

key performance indicators (KPIs) such as packet

transmission rates, latency, and network utilization.

OMNeT++ also allows researchers to implement cus-

tom result-handling mechanisms by deﬁning speciﬁc

statistical modules or writing output data directly to

external ﬁles. In this project, simulation results are

stored in .csv format, enabling easy post-processing

and visualization using external tools. By leverag-

ing OMNeT++’s ﬂexible data collection system, re-

searchers can efﬁciently analyse the impact of differ-

ent network conﬁgurations and mobility scenarios.

5 PERFORMANCE EVALUATION

In this section, we demonstrate MobiEdgeSim’s abil-

ity to simulate large-scale edge environments with

both static and mobile MEC servers. Speciﬁcally, we

evaluate how well the framework supports scalabil-

ity in node populations, dynamic resource constraints

under mobility, and different task-placement strate-

gies in realistic road trafﬁc conditions. Our goal is to

show that MobiEdgeSim effectively captures the per-

formance trade-off of MEC deployments when faced

with a variety of node types, workloads, and mobility

patterns.

5.1 Experimental Setup

Network Topology and Node Deployment. Our

simulation network is designed to replicate a realis-

tic urban environment by incorporating both mobile

and static nodes, enabling us to compare the impact

of MEC server mobility. Speciﬁcally, we draw in-

spiration from previous experimental setups (Kazmi

MobiEdgeSim: A Simulator for Large-Scale Mobile MEC Server Scenarios

Table 1: Evaluation conﬁgurations: Number of Static and

Mobile MEC Servers.

Conﬁguration Static Nodes Mobile Nodes Users

Conﬁg 1 100 0 100

Conﬁg 2 70 30 100

Conﬁg 3 300 0 100

Conﬁg 4 210 90 100

et al., 2025), adopting similar ranges for the num-

ber of nodes, mobility patterns, and trafﬁc condi-

tions. User Equipments (UEs) are randomly dis-

tributed throughout the simulation area, with their

mobility pattern generated using SUMO and man-

aged through Veins. For the static modules deploy-

ment, MEC servers are uniformly distributed across

the map. The MEC servers are deployed according

to four distinct conﬁgurations, as summarized in Ta-

ble 1:

• Conﬁguration 1. Comprises 100 static MEC

servers and 100 user equipments. This baseline

conﬁguration is used to assess the system’s per-

formance in the absence of server mobility.

• Conﬁguration 2. Incorporates a hybrid deploy-

ment consisting of 70 static MEC servers and 30

mobile MEC servers, along with 100 user equip-

ments. This setup introduces mobility and al-

lows investigation of dynamic resource allocation

strategies.

• Conﬁguration 3. Expands the static deployment

to 300 MEC servers, maintaining 100 user equip-

ments. This conﬁguration is designed to evalu-

ate the scalability of the system under large-scale

static infrastructure.

• Conﬁguration 4. Represents a large-scale hybrid

deployment featuring 210 static MEC servers and

90 mobile MEC servers. This conﬁguration com-

bines increased node density with signiﬁcant mo-

bility, enabling analysis of complex orchestration

in dynamic environments.

Table 2: Resource allocations for Users, Mobile MEC

Servers, and Static MEC Servers.

UE Static Server Mobile Server

RAM 2-16MB 1-10GB 0.5-5GB

Disk 20-160MB 1-2GB 0.5-1GB

CPU 10-160 MIPS 100-500 MIPS 50-250 MIPS

Resource Allocation. The resource parameters for

UEs and MEC servers are detailed in Table 2. Dur-

ing initialization, each node is assigned a maximum

resource capacity by randomly picking a value from

these ranges. Additionally, to better simulate a real-

istic workload environment, each node is initialized

with a randomly assigned baseline utilization ranging

from 10% to 90% of its total capacity. UEs are mod-

elled as resource-constrained devices, with memory

ranging from 2 MB to 16 MB, disk capacities be-

tween 20 MB and 160 MB, and CPU speeds from

10 MIPS to 160 MIPS. In contrast, MEC servers

are provisioned with substantially higher resources.

Speciﬁcally, static MEC servers receive memory be-

tween 1 GB and 10 GB, disk capacities from 1 GB

to 2 GB, and CPU speeds ranging from 100 MIPS

to 500 MIPS, while mobile MEC servers are allo-

cated slightly reduced resources—memory from 0.5

GB to 5 GB, disk capacities from 0.5 GB to 1 GB,

and CPU speeds between 50 MIPS and 250 MIPS.

These ranges are designed to reﬂect realistic hardware

constraints and allow our simulation to capture perfor-

mance variability, as each node’s resource capacity is

randomly determined within these bounds at runtime.

Placement Strategies. We evaluate three distinct task

placement strategies in this evaluation. Each strategy

attempts to select a suitable MEC server for service

placement. If no server meets the resource require-

ments, the placement request fails, and the task is de-

ferred until the MEC Orchestrator initiates the next

placement attempt.

The ﬁrst strategy, Random, selects an available

MEC server at random from the pool of candidates

with sufﬁcient resources. The second strategy, Best-

Fit (Hussein et al., 2019), systematically examines re-

source availability across all servers and selects the

one that best matches the task requirements by max-

imizing surplus resources. The third strategy, Clos-

estFit (Ouyang et al., 2018; Moubayed et al., 2020),

calculates the geographical distance between the user

and each MEC server, selecting the closest one that

can accommodate the task. In the following subsec-

tions, we assess the performance of these strategies

under varying conditions.

Bus with Mobile MEC Server

Mobile Users

Base Station

Static MEC Server

Figure 6: Conceptual overview of the simulated area in the

city of Dublin.

Simulation Area and Mobility Parameters. As

shown in Figure 6, our experiments take place in

a 12 km × 10 km region of central Dublin, derived

from an OpenStreetMap road network. We employ

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

the default urban driving model, which includes reg-

ular trafﬁc lights, speed limits, and potential conges-

tion at intersections. User devices are modelled as

cars following randomly generated routes, while mo-

bile MEC servers are mounted on buses that traverse

pre-deﬁned public transport routes. Vehicles adhere

to local road speed regulations and trafﬁc signals, but

their actual speed may be inﬂuenced due to real-time

trafﬁc conditions. By reﬂecting genuine street layouts

and multi-vehicle interactions, this setup captures re-

alistic urban dynamics and allows us to evaluate how

mobility-aware MEC scheduling strategies adapt un-

der city-scale scenarios.

5.2 Discussion of Performance

To validate the effectiveness and practicality of our

simulator, we performed experiments using the con-

ﬁgurations described in Section 5.1. Speciﬁcally,

these conﬁgurations were chosen to demonstrate how

MobiEdgeSim handles varying numbers of static and

mobile MEC servers, diverse resource allocations,

and user mobility patterns. During the experiments,

each user generates a service placement request once

per second, resulting in more than 4000 placement at-

tempts for each conﬁguration. By evaluating system

performance under these controlled and statistically

signiﬁcant conditions, our aim is to illustrate the capa-

bility of MobiEdgeSim to accurately model real-world

MEC challenges, including latency-sensitive work-

loads, resource bottlenecks, and dynamic server mo-

bility.

5.3 Resource Utilization of Static and

Mobile MEC Servers

Figure 7: Average and standard deviation of MEC server re-

source utilization across different scheduling strategies and

server conﬁgurations.

Figures 7 show the average and standard devia-

tion of resource utilization across different schedul-

ing strategies and server conﬁgurations. In all sce-

narios, utilization values appear similar with mini-

mal differences among scheduling strategies. We at-

tribute this consistency to the broad random baseline

resource utilization (10%–90%) assigned at simula-

tion start, which masks the impact scheduling poli-

cies would otherwise have. Moreover, our main goal

in this work is to demonstrate the effectiveness of

MobiEdgeSim in handling static and mobile MEC se-

tups rather than optimizing utilization (as the selected

placement models do not directly optimize the re-

source utilization). In future work, we intend to in-

clude the placement algorithms which can optimize

resources utilization and adopt to more diverse work-

load patterns to highlight the differences in schedul-

ing strategies and expose potential improvements in

utilization.

5.4 Key QoS Metrics: Distance and

Latency

To demonstrate how MobiEdgeSim measures essen-

tial QoS metrics in a mobile MEC environment, we

focus on two representative indicators: average user-

to-server distance and average latency. Speciﬁcally,

the distance metric refers to the geographic distance

calculated based on latitude and longitude coordinates

between each user equipment and its selected MEC

server. The latency metric represents the average net-

work delay, measured by calculating the transmission

time of timestamped UDP packets traveling between

user equipment and the assigned MEC server.

5.4.1 Distance

Figure 8 shows the average distance between a user

and its serving edge server under three scheduling al-

gorithms and four server conﬁgurations. Our simu-

lator records the real-time location of each user and

server, allowing us to compute distances when mak-

ing decision for each service placement request.

Figure 8: Average distance of connection between user

equipment and edge server with 95% conﬁdence intervals.

MobiEdgeSim: A Simulator for Large-Scale Mobile MEC Server Scenarios

Under the BestFit strategy, the average user-to-

server distance increases with server pool size and

server mobility, ranging from approximately 3,000m

in the smaller-scale 70S, 30M (green) scenario to

nearly 5,000m in the 300S conﬁguration. This trend

suggests that BestFit—while prioritizing resource

availability—can inadvertently select servers located

farther from the requesting user, especially as the total

server count grows. By contrast, the ClosestFit algo-

rithm consistently achieves minimal distances (gen-

erally under 1,000m) because it speciﬁcally aims to

choose the server with the shortest geographic sepa-

ration. Interestingly, conﬁgurations with large mobile

deployments 210S, 90M (blue) can yield slightly re-

duced distances under ClosestFit due to mobile nodes

being closer to users on average, yet the improve-

ments are modest if other factors (such as resource

constraints) come into play. Finally, with the Random

approach, distances tend to cluster in the mid-to-high

range (approximately 3,500m to 4,500m), reﬂecting

the lack of any proximity-based selection mechanism.

5.4.2 Latency

Figure 9: Average latency of connection between user

equipment and edge server with 95% conﬁdence intervals.

We also track the latency from a user to the selected

MEC server. Figure 9 presents the average latency

under the same four server conﬁgurations mentioned

above. By correlating location-based metrics with

resource utilization, MobiEdgeSim can reveal how

scheduling strategies adapt to changing network con-

ditions.

Results indicate that, with respect to latency, the

BestFit strategy typically achieves intermediate or rel-

atively low latency levels (around 200–320ms) across

all tested conﬁgurations. While this resource-centric

approach can sometimes result in larger physical dis-

tances (as observed in the BestFit distance metrics),

it generally avoids overloading any single server by

assigning tasks to whichever node currently appears

to have the most available resources. Such tran-

sient resource availability can result in subsequent

latency increases, particularly in large-scale, highly

dynamic environments. In contrast, ClosestFit ex-

hibits a broader range of latency outcomes, with the

210 static servers and 90 mobile servers conﬁgu-

ration reaching the highest average latency (about

500ms). This trend implies that ClosestFit, despite

minimizing geographic distance, risks imposing sig-

niﬁcant loads on geographically popular servers, cre-

ating queuing delays that negate the beneﬁts of prox-

imity. Meanwhile, the Random strategy falls some-

where in between (roughly 250–360ms), offering nei-

ther the distance-based beneﬁts of ClosestFit nor the

load-balancing advantage of BestFit. Its variability

further underscores the importance of employing in-

formed selection criteria in multi-access edge com-

puting, especially when numerous static and mobile

servers coexist.

Overall, these ﬁndings highlight a critical trade-

off between physical proximity and server load bal-

ancing. While ClosestFit can effectively minimize

distance, it may exacerbate congestion in heavily uti-

lized regions. Conversely, BestFit helps maintain rel-

atively even resource usage yet can increase propaga-

tion distance and occasionally make suboptimal as-

signments under transient spikes in demand. This

evaluation demonstrates that MobiEdgeSim can ac-

curately capture the complex interplay between ge-

ographic proximity, server mobility, and dynamic

resource allocation. The framework provides re-

searchers with a robust tool for systematically eval-

uating MEC deployment strategies under realistic ur-

ban conditions, facilitating deeper insights into how

these factors jointly inﬂuence end-to-end service la-

tency and performance.

6 CONCLUSIONS

In this paper, we presented MobiEdgeSim, an

OMNeT++-based simulation framework that inte-

grates Simu5G, Veins, and SUMO to support large-

scale dynamic scenarios involving both static and mo-

bile MEC servers. By combining detailed 5G New

Radio models, vehicular mobility simulations, and

realistic road trafﬁc conditions, our framework ad-

dresses a signiﬁcant gap in existing MEC simulators,

many of which overlook the complexities introduced

by server mobility.

Through our proof-of-concept experiments, we

have demonstrated that mobile MEC servers can ef-

fectively complement static edge nodes, enhancing

resource utilization and potentially reducing latency

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications

for real-time applications. Our comparative eval-

uations of different placement strategies (Random,

BestFit, and ClosestFit) reveal that resource-aware

and distance-aware algorithms can each yield unique

advantages depending on the network topology and

workload distribution. Notably, even a modest num-

ber of mobile MEC servers can signiﬁcantly improve

system performance in congested or high-density sce-

narios.

Our future work will primarily involve utilizing

MobiEdgeSim to explore more complex and diverse

application scenarios. This includes supporting richer

and more complex application scenarios, where a sin-

gle service may consist of multiple functions that can

be distributed across different edge servers to fur-

ther optimize performance. By simulating a wider

range of user requirements and more advanced server-

selection strategies (such as multi-objective heuris-

tics or learning-based methods), we aim to evalu-

ate how mobile MEC servers can best support next-

generation, latency-sensitive applications. By contin-

uing to reﬁne our platform and experimenting with

broader real-world use cases and orchestration at the

edge, we aim to provide deeper insights into how to

harness the synergy of static and mobile edge servers

in next-generation networks.

ACKNOWLEDGEMENTS

This publication has emanated from research con-

ducted with the ﬁnancial support of Taighde

Eireann – Research Ireland under Grant number

13/RC/2077 P2 at CONNECT: the Research Ireland

Centre for Future Networks, this work was also sup-

ported by VMware by Broadcom.

REFERENCES

Cabrera, C., Svorobej, S., Palade, A., Kazmi, A., and

Clarke, S. (2022). Maaco: A dynamic service place-

ment model for smart cities. IEEE Transactions on

Services Computing, 16(1):424–437.

Calheiros, R. N., Ranjan, R., Beloglazov, A., De Rose,

C. A., and Buyya, R. (2011). Cloudsim: a toolkit for

modeling and simulation of cloud computing environ-

ments and evaluation of resource provisioning algo-

rithms. Software: Practice and experience, 41(1):23–

50.

Gupta, H., Vahid Dastjerdi, A., Ghosh, S. K., and Buyya,

R. (2017). ifogsim: A toolkit for modeling and

simulation of resource management techniques in

the internet of things, edge and fog computing en-

vironments. Software: Practice and Experience,

47(9):1275–1296.

Hussein, M. K., Mousa, M. H., and Alqarni, M. A. (2019).

A placement architecture for a container as a service

(caas) in a cloud environment. Journal of Cloud Com-

puting, 8(1):7.

Kazmi, A. H., Staffolani, A., Zhang, T., Cabrera, C., and

Clarke, S. (2025). Dynamic service placement in

edge computing: A comparative evaluation of nature-

inspired algorithms. IEEE Access, 13:2653–2670.

Lera, I., Guerrero, C., and Juiz, C. (2019). Yafs: A simula-

tor for iot scenarios in fog computing. IEEE Access,

7:91745–91758.

Lopez, P. A., Behrisch, M., Bieker-Walz, L., Erdmann, J.,

otter

od, Y.-P., Hilbrich, R., L

ucken, L., Rummel, J.,

Wagner, P., and Wießner, E. (2018). Microscopic traf-

ﬁc simulation using sumo. In The 21st IEEE Interna-

tional Conference on Intelligent Transportation Sys-

tems. IEEE.

esz

aros, L., Varga, A., and Kirsche, M. (2019). Inet

framework. Recent Advances in Network Simula-

tion: The OMNeT++ Environment and its Ecosystem,

pages 55–106.

Moubayed, A., Shami, A., Heidari, P., Larabi, A., and Brun-

ner, R. (2020). Edge-enabled v2x service placement

for intelligent transportation systems. IEEE Transac-

tions on Mobile Computing, pages 1–1.

Nandan Jha, D., Alwasel, K., Alshoshan, A., Huang, X.,

Naha, R. K., Battula, S. K., Garg, S., Puthal, D.,

James, P., Zomaya, A. Y., et al. (2019). Iotsim-edge:

A simulation framework for modeling the behaviour

of iot and edge computing environments. arXiv e-

prints, pages arXiv–1910.

Nardini, G., Stea, G., Virdis, A., Sabella, D., et al. (2020).

Simu5g: a system-level simulator for 5g networks. In

Proceedings of the 10th International Conference on

Simulation and Modeling Methodologies, Technolo-

gies and Applications, pages 68–80. SciTePress.

Noferi, A., Nardini, G., Stea, G., and Virdis, A. (2023).

Rapid prototyping and performance evaluation of etsi

mec-based applications. Simulation Modelling Prac-

tice and Theory, 123:102700.

OpenStreetMap contributors (2017). Planet dump re-

trieved from https://planet.osm.org . https://www.

openstreetmap.org.

Ouyang, T., Zhou, Z., and Chen, X. (2018). Follow me at

the edge: Mobility-aware dynamic service placement

for mobile edge computing. IEEE Journal on Selected

Areas in Communications, 36(10):2333–2345.

Qayyum, T., Malik, A. W., Khattak, M. A. K., Khalid, O.,

and Khan, S. U. (2018). Fognetsim++: A toolkit for

modeling and simulation of distributed fog environ-

ment. IEEE Access, 6:63570–63583.

Riley, G. F. and Henderson, T. R. (2010). The ns-3 network

simulator. In Modeling and tools for network simula-

tion, pages 15–34. Springer.

Sommer, C., Eckhoff, D., Brummer, A., Buse, D. S., Hage-

nauer, F., Joerer, S., and Segata, M. (2019). Veins –

the open source vehicular network simulation frame-

work. In Virdis, A. and Kirsche, M., editors, Recent

Advances in Network Simulation. Springer.

MobiEdgeSim: A Simulator for Large-Scale Mobile MEC Server Scenarios

Sonmez, C., Ozgovde, A., and Ersoy, C. (2018). Edge-

cloudsim: An environment for performance eval-

uation of edge computing systems. Transac-

tions on Emerging Telecommunications Technologies,

29(11):e3493.

Varga, A. and Hornig, R. (2010). An overview of the om-

net++ simulation environment. ICST.

Zhang, T., Gallagher, O., Kazmi, A. H., and Clarke, S.

(2025). Mobiedgesim: A simulator for large-scale

mobile mec server scenarios. https://github.com/

MobiEdgeSim/MobiEdgeSim Cellular. Accessed:

2025-03-31.

SIMULTECH 2025 - 15th International Conference on Simulation and Modeling Methodologies, Technologies and Applications