TELK OMNIKA T elecommunication, Computing, Electr onics and Contr ol V ol. 24, No. 5, October 2026, pp. 1480 ∼ 1491 ISSN: 1693-6930, DOI: 10.12928/TELK OMNIKA.v24i5.27849 ❒ 1480 V irtualized e v olv ed pack et cor e deployment strategies in emer ging P alestinian mobile netw orks Saed T arapiah 1 , Shadi Atalla 2 , Khawla Haz Daraghme 3 1 T elecommunication Engineering Department, An–Najah National Uni v ersity , Nablus, P alestine 2 Colle ge of Engineering and Information T echnology , Uni v ersity of Dubai, Dubai, United Arab Emirates 3 Ooredoo P alestine, Ramallah, P alestine Article Inf o Article history: Recei v ed Feb 3, 2026 Re vised Jun 21, 2026 Accepted Jun 30, 2026 K eyw ords: Emer ging mobile netw orks Deplo yment strate gy KPI-based weighted scoring Mobile core netw ork virtualization Multi-criteria decision analysis Netw ork function virtualization V irtualized e v olv ed pack et core ABSTRA CT Mobile operators in emer ging mark ets f a ce rising data demand under re v enue, infrastructure, re gulatory , and in v estment constraints. Netw ork function vir - tualization (NFV) enables e v olv ed pa ck et core (EPC) modernization, b ut op- erators must determine which deplo yment model pro vides the best strate gic and operational t. This paper e xamines EPC virtual ization as an operator - le v el deplo yment-selection problem using P alestine as a case study . A conte xt- specic decision frame w ork inte grates operator readiness, stak eholder -deri v ed k e y performance indicator (KPI) weights, deplo yment-model ratings, capital e xpenditure (CapEx) / operating e xpenditure (OpEx) analysis, anal ytic hierar - ch y process (AHP) consistenc y checking, and technique for order of preference by similarity to ideal solution (T OPSIS)-style rob ustness v alidation. Fi v e al- ternati v es are compared under the same operator scenario: o wned le g ac y EPC, o wned NFV / virtualiz ed e v olv ed pack et core (vEPC), netw ork functions virtual- ization infrastructure as a service (NFVIaaS), virtual netw ork platform as a ser - vice (VNP aaS), and virtual netw ork service as a service (VNSaaS). The no v elty lies in linking multi-criteria scoring to local re gulatory , infrastructure, security , supply-chain, and b usiness constraints rather than applying generic weighted ranking alone. Scenario-based technical v alidation e xamines latenc y / quality of service (QoS) e xposure, scalability , reco v ery , elasticity , and observ ability . The results indicate that fully o wned vEPC pro vides the strongest o v erall t when service control, quality , implementation constraints, and long-term operational ef cienc y are jointly considered. This is an open access article under the CC BY -SA license . Corresponding A uthor: Saed T arapiah T elecommunication Engineering Department, An–Najah National Uni v ersity Nablus, P alestine Email: s.tarapiah@najah.edu 1. INTR ODUCTION Mobile operators in emer ging mark ets must e xpand mobile broadband, internet of things (IoT), machine- to-machine (M2M) , cloud, and lo w-latenc y services while f acing re v enue pressure, spectrum limitations, in- frastructure restrictions, and limited in v estment capacity . Le g ac y enabl es e v olv ed pack et core (EPC) deplo y- ments, which are tightly coupled to proprietary hardw are, are therefore increasingly dif cult to scal e economi- cally . Netw ork function virtualization (NFV) of fers a modernization path by decoupling mobile-core functions from dedicated appliances and supporting virtualized or cloud-based deplo yment of EPC functions [1]-[5]. Re- J ournal homepage: http://telk omnika.uad.ac.id Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control ❒ 1481 cent w ork conrms the importance of softw are-dened netw orking (SDN) / NFV inte gration, cloud-nati v e core functions, service-based architectures, and virtualized core in v estment for 5G e v olution [6]-[10]. The central problem for constrained operators is not only whether EPC virtualization is feasible, b ut which deplo yment and service model should be selecte d under technical, b usiness, re gulatory , and operational constraints. This issue is critical in P alestine, where operators f ace infrastructure restrictions, supply-chain lim- itations, re gulatory uncertainty , and e xternal competition. Existing studies mostly optimize or v alidate selected virtualized/cloud-nati v e architectures through resource allocation, t estbeds, observ ability , sl ice federa tion, se- curity , or orchestration [11]-[18]. Fe wer studies pro vide an operator -le v el decis ion frame w ork that compares alternati v e EPC deplo yment models under one emer ging-mark et scenario and translates the ranking into cost, go v ernance, security , re gulatory , and operational-control implications. Accordingly , the research question is: Whic h EPC deployment model pr o vides the most suitable bal- ance of cost, quality , time to mark et, securit y , r e gulatory alignment, value pr oposition, and oper ational com- ple xity for a P alestinian mobile oper ator oper ating under emer ging-mark et constr aints? The no v elty is not a ne w generic multi-criteria decision analysis (MCD A) algorithm; rather , it is the adaptation of multi-criteria scoring into a deplo yment-specic EPC virtualization decision frame w ork that links stak eholder priorities, an- alytic hierarch y process (AHP)-supported k e y performance indicator (KPI) weighting, technique for order of preference by similarity to ideal solution (T OPSIS)-style rob ustness checking, total cost of o wnership (TCO) estimation, and scenario-based technical v alidation to local operating constraints. The main contrib utions are: (i) formulating EPC virtualization as a deplo yment-selection problem; (i i) proposing a conte xt-a w are decision frame w ork; (iii) comparing o wned le g ac y EPC, o wned NFV / virtualized e v olv ed pack et core (vEPC), netw ork functions virtualization infrastructure as a service (NFVIaaS), virtual netw ork platform as a service (VNP aaS), and virtual netw ork service as a service (VNSaaS) with unied KPIs; (i v) clarifying ho w this w ork dif fers from NFV studies focused mainly on technical optimizati on or testbed v alidation; and (v) sho wing wh y o wned vEPC can outperform lo wer -capital e xpenditure (CapEx) outsourced models when quality of service (QoS) go v ernance, operational control, security accountability , and long-term operating e xpenditure (OpEx) are jointly considered. 2. COMPREHENSIVE THEORETICAL B ASIS AND RELA TED W ORK 2.1. Legacy EPC and virtualized EPC backgr ound The EPC is the internet protocol (IP)-based mobile-core architecture responsible for mobility man- agement, session control, po l ic y enforcement, subscriber data handling, and pack et-g ate w ay functions. Its main logical elements include the mobility management entity (MME), serving g ate w ay (SGW), pack et data netw ork g ate w ay (PGW), polic y and char ging rules function (PCRF), and home subscriber serv er (HSS) [19]. In con v entional operator deplo yments, these functions are usually implemented through appliance-centric and v endor -specic platforms. Such platforms are stable and operationally f amiliar , b ut the y are less adaptable to f ast traf c gro wth, service di v ersication, and rapi d capacity e xpansion. Figure 1 illustrates the con v entional EPC logical structure used as the baseline in this study . Figure 1. Con v entional le g ac y EPC architecture used as the baseline deplo yment model NFV changes this model by decoupling netw ork functions from dedicated hardw are and enabling EPC functions to run as VNFs o v er a shared NFV infrastructure (NFVI), go v erned by management and orchestration functions [5], [20]. This separation allo ws operators to pool compute resources, automate lifec ycle operations, scale VNFs according to traf c demand, and reduce dependence on single-purpose appliances. Ho we v er , the V irtualized e volved pac k et cor e deployment str ate gies in emer ging P alestinian ... (Saed T ar apiah) Evaluation Warning : The document was created with Spire.PDF for Python.
1482 ❒ ISSN: 1693-6930 v alue of NFV depends strongly on the selected deplo yment model. A model with lo w ini tial CapEx may intro- duce e xternal dependenc y , weak er observ ability , stricter service le v el agreement (SLA) reliance, or re gulatory e xposure; con v ersely , an o wned model may require higher initial in v estment b ut pro vide stronger long-term control. 2.2. Deployment models consider ed in this study Fi v e deplo yment alternati v es are e v aluated under one x ed operator scenario. The le g ac y EPC model represents the baseline appliance-based architecture. Owned NFV/vEPC means that the operator o wns and manages the NFVI, VNFs, and lifec ycle-control processes. NFVIaaS shifts the infrastructure layer to an e xter - nal pro vider while k eeping VNF control lar gely with the operator . VNP aaS pro vides an inte grated virtualized platform on which netw ork services can be depl o yed. VNSaaS pro vides the netw ork softw are as a service and therefore shifts the greatest portion of control to the service pro vider . T able 1 summarizes the decision rele v ance of these models. T able 1. Deplo yment models and main decision implications Model Ownership and control Main benet Main risk Le g ac y EPC Operator o wns appliance-based EPC F amiliar operation and direct con- trol W eak elasticity and high lifec ycle cost Owned NFV/vEPC Operator o wns NFVI, VNFs, and orchestration Strong control, scalability , and ob- serv ability Higher implementation and skill re- quirements NFVIaaS Pro vider o wns infrastructure; oper - ator manages VNFs Reduced CapEx and infrastructure b urden Pro vider dependenc y and cross- domain troubleshooting VNP aaS Pro vider of fers virtualized platform F aster service de v elopment and platform abstraction Reduced control o v er platform be- ha vior VNSaaS Pro vider of fers netw ork softw are as a service Minimal initial CapEx and rapid consumption Highest e xternal dependenc y and recurring OpEx 2.3. P ositioning against existing w ork Recent NFV and fth generation (5G)-core literature can be grouped into four streams. The rst stream focuses on cloud-nati v e resource allocat ion and automati o n for virtualize d or microservic e-based cores [11]. The second de v elops e xperimental platforms and testbeds for 5G-core deplo yment, analytics, and observ- ability [12], [14]. The third addresses slice federation, multi-domain orchestration, and cross-administrati v e- domain service continuity [13]. And the fourth studies ener gy ef cienc y , user plane function (UPF) placement, security , and resilience in softw arized cores [15]-[18]. These w orks impro v e important mechanisms after a core architecture has already been selected. The present study addresses a dif ferent decision le v el. It does not propose a ne w generic opti miza- tion algorithm or a single testbed implementation. Instead, it formalizes the earlier operator decision: which EPC deplo yment model should be s elected before i mplementation under constrained mark et conditions. This requires simultaneous consideration of technical feasibility , TCO, re gulatory e xposure, security accountabil- ity , troubleshooting autonomy , service control, and pro vider dependence. T able 2 positions the study ag ainst representati v e recent w ork. T able 2. No v elty positioning relati v e to representati v e recent NFV/5G-core w ork Approach Prima ry output Dif ference of this study Cloud-nati v e resource allocation [11] Allocation polic y or resource optimization This w ork selects among deplo yment/service models be- fore resource-optimization policies are applied 5G-core testbeds [12] Experiment al deplo yment and measurements This w ork pro vides a decision frame w ork; testbeds are fu- ture v alidation instruments for the selected model Slice federation [13] Cross-dom ain slice orchestration This w ork determines the baseline EPC/vEPC o wnership and service model before slice federation is considered Ener gy/UPF studies [15], [16] Placement and ener gy-performance tradeof f This w ork maps ener gy and performance implications into operator -le v el KPIs and TCO interpretation Observ ability studies [14] Monitoring and troubleshooting mechanisms This w ork treats observ ability as a decision criterion af- fecting model selection and go v ernance This positioning claries the contrib ution: the weighted scoring, AHP , and T OPSIS components are not presented as ne w algorithms. The y are used as formal instruments inside a conte xt-a w are EPC deplo yment- selection frame w ork. The no v elty is therefore methodological and practical: the frame w ork translates the broad TELK OMNIKA T elecommun Comput El Control, V ol. 24, No. 5, October 2026: 1480–1491 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control ❒ 1483 NFV modernization problem into an auditable operator decision process for a constrained emer ging-mark et en vironment. 3. METHOD This study uses a case-study-based MCD A design. The scenario is x ed for al l alternati v es in t erms of EPC functi onal scope, traf c-gro wth assumptions, re gulatory setting, service portfolio, redundanc y tar get, and mark et constraints. Empirical input w as obtained through structured and semi-structured discussions with an anon ymized panel of se v en e xperts selected for direct in v olv ement in mobile-core pla n ni ng, EPC operation, NFV / cloud infrastructure, QoS, security / re gulatory compliance, or telecom b usiness strate gy are summarized in T able 3. The process had three st ages: conrmation of the v e deplo yment alternati v es, re vi e w of the KPI set, and comparati v e judgments for KPI weighting and model scoring. Aggre g ated group judgment w as used rather than a single-respondent score. T able 3. Anon ymized e xpert panel used for KPI elicitation Background No. Ev aluation role Mobile-core / EPC engineering 2 EPC functions, migration feasibility , control / user -plane implications NFV / cloud infrastructure 1 NFVI, orchestration, virtualization readiness, scalability Netw ork operations / QoS 2 A v ailability , SLA assurance, troubleshooting, service continuity Security / re gulatory / go v ernance 1 Compliance, security e xposure, data control, re gulatory constraints Business strate gy / procurement 1 CapEx/OpEx, v endor dependenc y , time to mark et, b usiness v alue T otal 7 Aggre g ated e xpert judgment for weights and scores AHP is used to formalize e xpert judgments and reduce subjecti vity . F or comparison matrix A = [ a ij ] , the weight v ector is obtained from: Aw = λ max w , C I = λ max − n n − 1 , C R = C I R I . (1) Judgments with unacceptable inconsistenc y should be re vie wed before nalizing the weights [21], [22]. The nal KPI weight is calculated as (2). w k = 0 . 25 b k + 0 . 15 t k + 0 . 60 d k , (2) where b k , t k , and d k are normalized AHP-supported contrib utions from b usiness-strate gy , transformation- trigger , and technical-deplo yment vie wpoints. The coef cients reect the case-specic importance of technical feasibility , security , operational control, and service continuity , while still incorporating b usiness and transfor - mation dri v ers. Each model m is rated ag ainst KPI k using r m,k ∈ [1 , 5] and scored as (3). S m = P K k =1 w k r m,k P K k =1 w k (3) A T OPSIS-style rob ustness check is then applied using a normalized matrix v m,k = w k r m,k / q P M m =1 r 2 m,k , positi v e/ne g ati v e ideals A + and A − , distances D + m and D − m , and the closeness coef cient: C m = D − m D + m + D − m (4) A higher C m means the model is closer to the positi v e ideal and f arther from the ne g ati v e ideal [23]. T o preserv e reproducibility , the scoring matrix w as constrained by four controls. First, all models were e v aluated using the same EPC functional scope and the same planning horizon. Second, each KPI w as inter - preted using a predened direction: higher v alues indicate stronger suitability , e xcept for cost and comple xity where lo wer e xposure is con v erted into a higher suitability score. Third, e xpert disagreement w as treated as a trigger for clarication, not as an automatic a v erage when meanings dif fered. And fourth, cost, go v ernance, security , and technical-performance e vidence were cross-check ed ag ainst the literature to a v oid relying solely on e xpert perception. These controls are summarized in T able 4. V irtualized e volved pac k et cor e deployment str ate gies in emer ging P alestinian ... (Saed T ar apiah) Evaluation Warning : The document was created with Spire.PDF for Python.
1484 ❒ ISSN: 1693-6930 T able 4. Reproducibility and v alidity controls in the e v aluation frame w ork Control Purpose Fix ed operator scenario Ensures that dif ferences among alternati v es come from the deplo yment model, not from changing demand or scope assumptions Common KPI direction Con v erts cost, comple xity , security , and QoS criteria into comparable suitability scores Aggre g ated e xpert judgment Reduces single-respondent bias and forces clarication when interpretations di v er ge AHP consistenc y check Detects inconsistent pairwise judgments before weights are nalized T OPSIS rob ustness check T ests whether the preferred model remains stable under ideal-solution interpretation TCO and scenario v alidation Adds nancial and technical support be yond subjecti v e scoring 4. RESUL TS AND DISCUSSION 4.1. Case-study r eadiness and KPI priorities The case-study assessment indicates that W ataniya Mobile has strate gic interest in EPC virtuali zation because the transformation is link ed to service agility , long-term cost cont rol, impro v ed QoS go v ernance, and preparation for future 4G/5G-oriented core e v olution. The operator conte xt also sho ws that virtualization cannot be e v aluated only as a technology-upgrade project. It must be e v aluated as a combined technical, nancial, re gulatory , and operational-control decision. Stak eholder input emphasized that OpEx reduction and QoS/SLA assurance are the strongest decision dri v ers. This is consist ent with constrained-mark et economics, where traf c gro wth increases operating pres- sure while re v enue per unit of traf c may decline. Re gulatory-constraint mitig ation, operational-comple xity reduction, and security control were also prioritized because outsourced models can shift control points outside the operator boundary . T able 5 summarizes the condensed KPI hierarch y used in the nal e v aluation. T able 5. Condensed KPI priority structure used in the case study Rank KPI group Decision meaning in the P alestinian operator conte xt 1 OpEx reduction Long-term sustainability , lifec ycle cost control, and operational automation 2 QoS, a v ailability , and SLA assurance Ability to maintain service quality under gro wth, f ailures, and traf c peaks 3 Re gulatory-constraint mitig ation Fit with local infrastructure limitations, polic y uncertainty , and data-control needs 4 Operational-comple xity reduction Ease of troubleshooting, lifec ycle control, process adaptation, and staf f readiness 5 Security-threat minimization Control o v er sensiti v e core functions, interf aces, data e xposure, and attack surf ace 6 T ime-to-mark et and v alue proposition Ability to launch ne w services and support digital-transformation objecti v es 7 CapEx and implementation comple xity Initial af fordability , procurement feasibility , deplo yment risk, and migration ef fort The priority structure conrms wh y the preferred model should not be selected using CapEx alone. The operator needs a model that balance s cost, control, QoS, and go v ernance. Therefore, t he e v aluation gi v es strong weight to the technical-deplo yment vie wpoint while retaining b usiness-strate gy and transformation- trigger dimensions through the AHP-supported weighting model. 4.2. KPI, cost, and deployment-model e v aluation The nal KPI set co v ers OpEx, QoS/a v ailability/SLA assurance, re gulatory-constraint mitig ati on , op- erational comple xity , security control, ti me to mark et, v alue proposition, CapEx, management comple xity , implementation comple xity , and re gulatory compliance. The cost component is modeled as v e-year TCO. F or deplo yment model m , C apE x m = C H W m + C S W m + C N F V I m + C M AN O m + C I N T m + C M I G m + C T R m + C C O N T m , (5) O pE x m = Y X y =1 C S U P m,y + C LI C m,y + C E N E m,y + C F AC m,y + C S T AF F m,y + C S LA m,y + C N E T m,y + C O P S m,y T C O m = C apE x m + O pE x m (6) F or service-based alternati v es, recurring service cost is represented by (7). C S V C m,y = F m + α m D y + β m R y + γ m A y , (7) where F m is the x ed annual char ge, D y is capacity demand, R y is redundanc y le v el, and A y is the SLA class. Cost assumptions com bine operator planning inputs, aggre g ated commercial estimates, NFV architecture TELK OMNIKA T elecommun Comput El Control, V ol. 24, No. 5, October 2026: 1480–1491 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control ❒ 1485 assumptions, e xpert judgment, ener gy/performance literature, and mark et benchmarks [15], [24]-[26]. Public mark et reports are used as secondary trend checks, not direct price inputs [9], [10]. T able 6 summarizes the major cost components and sources used to mak e the calculation reproducible without disclosing condential v endor quotations. T able 6. Cost model assumptions and estimation basis Cost element Included items Estimation basis Hardw are/NFVI CapEx EPC appliances, commodity serv ers, storage, switching, virtualization layer Operator planning input, aggre g ated v endor esti- mates, NFV architecture assumptions [5], [24] Softw are/management and orchestration (MANO) CapEx EPC/vEPC licenses, VNF packages, orchestra- tion, inte gration middle w are NFV deplo yment literature and operator technical assumptions [5], [6] Migration/training Design, testing, migration planning, accep- tance, staf f training Expert-panel judgment and transformati on assump- tions [27] Recurring licenses/support Subscriptions, managed-service fees, hard- w are/softw are support Aggre g ated commercial estimates and mark et benchmarks [9], [10] Ener gy/f acilities/operations Po wer , cooling, oor space, monitoring, backup, patching, troubleshooting Ener gy/performance and lifec ycle-management lit- erature [15], [26] SLA/pro vider dependenc y Connecti vity , redundanc y premiums, cross- domain support, e xternal escalation Service-model assumptions and e xpert judgment [24], [25] Figure 2 visualizes the cost and nal-score results. The visual comparison sho ws that servi ce-based options reduce initial CapEx, whereas o wned NFV/vEPC pro vides the strongest nal s core and the lo west v e-year OpEx as sho wn in T able 7 and Figures 2(a) and (b), while T able 8 summarizes the comparati v ee v alu- ation of each deplo yment model across the complete decision KPI set. The r esults in T able 7 sho w that o wned NFV/vEPC has the highest o v erall score and the lo west OpEx, although it is not the lo west-CapEx option. NFVIaaS and VNP aaS are attracti v e for capital-constrained scenarios, b ut the y introduce stronger pro vider de- pendenc y and cross-domain troubleshooting/SLA e xposure. VNSaaS minimizes initial CapEx b ut accumulates high recurring OpEx and weak ens direct EPC performance go v ernance. Le g ac y EPC remains f amiliar and controlled b ut lacks elasticity and produces the weak est o v erall case t. (a) (b) Figure 2. Cost and nal KPI-score results for: (a) v e-year TCO structure and (b) nal KPI score ranking V irtualized e volved pac k et cor e deployment str ate gies in emer ging P alestinian ... (Saed T ar apiah) Evaluation Warning : The document was created with Spire.PDF for Python.
1486 ❒ ISSN: 1693-6930 T able 7. Fi v e-year CapEx, OpEx, TCO, and nal core Model CapEx (K USD) OpEx (K USD) TCO (K USD) Score Le g ac y EPC 30,097 18,118 48,215 2.4 Owned NFV/vEPC 21,016 14,950 35,966 3.7 NFVIaaS 14,353 18,339 32,692 3.4 VNSaaS 0 39,035 39,035 3.1 VNP aaS 1,211 31,703 32,915 3.1 T able 8. Summary comparison of deplo yment models across decision KPIs KPI Le g ac y Owned vEPC NFVIaaS VNP aaS VNSaaS OpEx reduction L VH M M L QoS/SLA assurance M VH H M M Re gulatory mitig ation L H M M L–M Operational comple xity L H H M M Security control H H M M L–M T ime to mark et M VH H M M V alue proposition L H M H H CapEx reduction L M H VH VH Management comple xity L M M H H Implementation comple xity L M M–H H H Re gulatory compliance M H M M L–M Ov erall case t L VH H M M VH: v ery high, H: high, M: medium, and L: lo w . Ratings denote relati v e suitability under the same case scenario 4.3. Scenario-based technical v alidation Because the ranking includes performance-related claims, the results are complemented by s cenario- based technical v alidation under normal-load, peak-l oad, and stress-load conditions as sho wn in T able 9. End- to-end delay is decomposed as (8). L e2e = L pro c + L queue + L trans + L ctrl , (8) where the terms represent processing, queuing, transport, and control/orchestration delay . Scalability is inter - preted through an elasticity indicator: E m = ∆ C m ∆ T m + ∆ O m , (9) where ∆ C m is additional capacity , ∆ T m is time to acti v ate capacity , and ∆ O m is additional operational ef fort or dependenc y . Owned vEPC can reduce queuing pressure through local VNF scaling while preserving direct function-placement, monitoring, and lifec ycle control. NFVIaaS also of fers elasticity , b ut pro vider dependence may af fect transport delay , SLA enforcement, and root-cause visibility . This v alidation does not claim li v e-netw ork measurement; rather , it pro vides an analytical performance basis for the case-study ranking. It supports the conclusion that o wned vEPC is preferred not only because of nancial scoring b ut also because it pro vides the most coherent balance of latenc y go v ernance, elastic scaling, reco v ery control, and observ ability . A li v e pilot remains necessary before commercial rollout. T able 9. Condensed technical v alidation matrix Dimension Le g ac y Owned vEPC NFVIaaS VNP aaS VNSaaS Latenc y/QoS Stable b ut rigid Best local control Good, pro vider af fected Platform con- strained SLA/service depen- dent Peak-load scaling Hardw are-bound High elasticity High b ut e xternal Medium Medium Reco v ery Slo w appliance reco v ery VNF migration/l ocal orchestration Pro vider - dependent Platform- dependent Service-dependent Observ ability V endor -specic Full local visibility Interf ace-limited Abstracted Most abstracted Utilization Lo w pooling High pooling/control High pooling Medium–high Pro vider -managed 4.4. Rob ustness inter pr etation and sensiti vity analysis A qualitat i v e sensiti vity analysis w as added to test whether the conclusion depends e xcessi v ely on a single KPI group. Three interpretation scenarios were e xamined. The rst is the base case, where technical TELK OMNIKA T elecommun Comput El Control, V ol. 24, No. 5, October 2026: 1480–1491 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control ❒ 1487 deplo yment has the highest inuence because the operator must preserv e service control and continuity . The second is a cost-constrained case, where CapEx pressure is stronger and outsourced models ar e e xpected to benet. The third is a go v ernance-constrained case, where security , re gulatory e xposure, observ ability , and troubleshooting control dominate. T able 10 summarizes the e xpected ranking stability . The sensiti vity interpretation sho ws that the o wned vEPC recommendation is rob ust when the deci- sion emphasizes long-term control, service quality , security , re gulatory accountability , and OpEx optimization. Ho we v er , the ranking is not uni v ersal. If a dif ferent operator has v ery limited capital, strong trust in an NFVI pro vider , mature cloud re gulation, and less st rict control requirements, NFVIaaS or VNP aaS may become more competiti v e. This reinforces the main purpose of the proposed frame w ork: it is not a x ed ranking tool, b ut a transparent decision process that can be recalibrated for other operators and mark ets. T able 10. Sensiti vity interpretation of the deplo yment-model ranking Scenario Expected strongest model Interpretation Base-case weighting Owned NFV/vEPC Balanced control, OpEx reduction, QoS go v ernance, and technical t k eep o wned vEPC rst CapEx-emphasis scenario NFVIaaS or VNP aaS Outsourced or platform-based models impro v e when initial in v estment dominates the decision Go v ernance-emphasis scenario Owned NFV/vEPC Local control, observ ability , compliance, and security ac- countability strengthen o wned vEPC Rapid-service scenario VNP aaS or VNSaaS Service-based models impro v e when f ast service consump- tion is prioritized o v er deep core control Stress-load scenario Owned NFV/vEPC or NFVIaaS Elastic scaling is important; o wned vEPC retains stronger visibility and reco v ery control 4.5. General discussion The ranking should be interpreted in the economics of constrained mobile mark ets, where OpEx con- trol, QoS continuity , local accountability , and re gulatory risk are central to architecture choice. The frame- w ork’ s practical v alue is that it mak es e xplicit wh y a lo w-CapEx outsourced model may be less suitable when data control, security responsibility , troubleshooting autonomy , supply-chain e xposure, and long-term go v er - nance are important. It therefore dif fers from studies that report resource-allocation, slicing, or testbed perfor - mance g ains after a core architecture has already been selected. The result also sho ws that cost should not be interpreted only through initial capital e xpenditure. VNSaaS and VNP aaS reduce upfront o wnership cost, b ut their recurring service fees, SLA premiums, and pro vider -dependent opera tions increase long-term e xposure. NFVIaaS of fers a stronger balance for capital- constrained operators, b ut it still transfers infrastructure visibility and some reco v ery responsibilities to the pro vider . Owned NFV/vEPC is less attracti v e when only short-term CapEx is considered, yet it becomes the strongest option when service control, observ ability , reco v ery , security accountability , and OpEx optimization are included. F or the P alestinian case , re gulatory uncertainty and infrastructure restrictions increase the v alue of local control. The preferred model should therefore support technical modernization without creating unneces- sary dependence on e xternal platforms or opaque SLA enforc ement. This e xplains wh y o wned vEPC emer ges as the strongest case t, while outsourced service models remain useful as partial, transitional, or h ybrid options rather than as full replacements for operator -controlled core go v ernance. 4.6. Practical implications f or emer ging-mark et operators The proposed frame w ork has se v eral practical i mplications be yond the W ataniya Mobile case. First, i t pro vides a structured w ay to separate short-term af fordability from long-term suitability . Operators often com- pare deplo yment models primarily through procurement cost, b ut this may conceal recurring service char ges, escalation delay , platform dependenc y , and loss of observ ability . The TCO and technical-v alidation layers help re v eal these hidden dimensions. Second, the frame w ork highlights that the migration path can be phased. A constrained operator does not need to mo v e directly from le g ac y EPC to a full cloud-nati v e core. A practical roadmap may be gin with a limited vEPC proof of concept, continue with partial virtualization of selected EPC functions, and then e xpand to w ard broader NFV/MANO inte gration after staf f skills, operational processes, monitoring tools, and security controls mature. This phased interpretation reduces implementation risk while preserving the long-term tar get architecture. V irtualized e volved pac k et cor e deployment str ate gies in emer ging P alestinian ... (Saed T ar apiah) Evaluation Warning : The document was created with Spire.PDF for Python.
1488 ❒ ISSN: 1693-6930 Third, the model can support e x ecuti v e communication. T echnical teams usually e v aluate sca lability , redundanc y , reco v ery , and observ ability , while b usiness teams focus on CapEx, OpEx, v endor contracts, and time to mark et. Re gulatory and security teams focus on data control, compliance, and accountability . By placing these concerns in one matrix, the frame w ork allo ws decision mak ers to see ho w the selected model performs across multiple or g anizational vie wpoints rather than treating the choice as a purely technical or purely nancial decision. F ourth, the frame w ork can be used as a migration-control instrument. The operator can dene decision g ates before each imple mentation phase. At each g ate, the team can rea ssess technical readiness, cost e xposure, v endor dependenc y , re gulatory constraints, security posture, and measurable service quality . This is especially important in emer ging mark ets because procurement delays, infrastructure restrictions, and cloud-readiness g aps may change the feasibility of the selected model during i mplementation. T able 11 pro vides a compact roadmap that k eeps the paper practical while a v oiding unsupported claims of a completed pilot deplo yment. T able 11. Suggested phased roadmap for applying the selected vEPC model Phase Main acti vity Decision g ate Readiness audit Re vie w EPC assets, skill g aps, monitoring tools, and re gulatory constraints Conrm that virtualization scope and risk boundaries are clear Proof of concept Deplo y limited vEPC functions in a controlled lab or isolated en vironment V alidate functionality , orchestration beha vior , and op- erational visibility Controlled eld trial T est selected functions under limited li v e- traf c conditions Measure latenc y , a v ailability , reco v ery , and trou- bleshooting performance Gradual migration Expand virtualization to addit ional functions or traf c domains Conrm that SLA, security , and operational processes remain stable Optimization T une scaling policies, resource allocation, cost controls, and monitoring Recalibrate KPI weights and update the model ranking if conditions change This roadmap is als o useful for pre v enting o v er generalizati on. The conclusion that o wned vEPC is preferred in this study should be treated as a case-specic decision result support ed by cost modelling, e xpert judgment, rob ustness interpretation, and scenar io-based v alidation. It should not be read as a uni v ersal claim that all emer ging-mark et operators must adopt the same model. The frame w ork remains transferable, b ut the weights, cost v al ues, pro vider assumptions, and re gula tory inputs must be recalibrated for each operator sho wn in T able 12. T able 12. Practical use of the frame w ork by operator stak eholders Stak eholder group Ho w the frame w ork supports the decision Netw ork engineering Compares scalability , latenc y e xposure, redundanc y , and function-placement control Operations and QoS teams Ev aluates observ ability , troubleshooting visibility , SLA assurance, and reco v ery beha vior Security and re gulatory teams Examines data control, pro vider dependenc y , compliance e xposure, and accountability boundaries Finance and procurement Compares CapEx, recurring OpEx, TCO, SLA premiums, and long-term v endor dependenc y Ex ecuti v e management Links the technical architecture decision to strate gic control, in v estment risk, and service dif ferenti- ation 4.7. Limitations The study remains a case-study-based decision analysis and should not be interpreted as a uni v er - sal ranking of EPC deplo yment models. The results depend on the P alestinian operator conte xt, the selected KPI set, the e xpert-panel judgments, and the v e-year TCO assumptions. In mark ets with mature cloud re g- ulation, stronger national NFVI pro viders, lo wer connecti vity risk, or dif ferent traf c proles, NFVIaaS or platform-based models may recei v e s tronger rankings. In addition, the technical v alidation is scenario-based and analytical; it supports the interpretation of the ranking b ut does not replace li v e measurements. These limitations are addressed by the future pilot-v alidation path proposed belo w . 4.8. Futur e w ork Future w ork should v alidate the r anking through staged pilot deplo yments, be ginning with a limited vEPC proof of concept and controlled eld trials. Measured indicators should include latenc y , jitter , pack et loss, throughput, a v ailability , VNF instantiation time, s caling time, reco v ery time, troubleshooting duration, resource utilization, and ener gy consumption [12], [14], [15]. The measured v alues can then be compared with the predicted KPI scores to recalibrate the weighting and scoring matrix. TELK OMNIKA T elecommun Comput El Control, V ol. 24, No. 5, October 2026: 1480–1491 Evaluation Warning : The document was created with Spire.PDF for Python.
TELK OMNIKA T elecommun Comput El Control ❒ 1489 A second e xtension is to model edge-enabled EPC and 5G-core placement. Future v ersions should e v aluate the placement of control-plane and user -plane functions across central cloud, re gional cloud, and edge sites. This w ould allo w t he frame w ork to account for local break out, transport cost , edge-site a v ailability , orchestration o v erhead, ener gy footprint, and UPF/vUPF placement decisions [16], [28]. A third e xtension is multi-operator slicing and infrastructure sharing. In emer ging mark ets, shared NFVI, neutral-host deplo yments, and cross-operator slice federation may become attracti v e because of spec- trum scarcity , capital limitations, and infrastructure restrictions. Extending the frame w ork in thi s direction requires additional KPIs for slice isolation, inter -operator go v ernance, data o wnership, cross-domain SLA as- surance, re gulatory compliance, and re v enue-sharing mechanisms [13], [17], [18]. 5. CONCLUSION This paper presented a compact, conte xt-a w are decision frame w ork for selecting EPC virtualiza tion strate gies in an emer ging P alestinian mobile-netw ork scenario. The frame w ork combines e xpert-based KPI elicitation, AHP-supported weighting, weighted scoring, T OPSIS-style rob ustness checking, v e-year TCO estimation, and scenario-based technical v alidation. The analysis indicates that o wned NFV/vEPC pro vides the strongest o v erall t for W ataniya Mobile when OpEx reduction, QoS/SLA assurance, operational control, security go v ernance, re gulatory e xposure, implementation constraints, and long-term service e xibility are considered together . NFVIaaS and VNP aaS remain useful alternati v es for operators prioritizing lo wer initial CapEx, b ut their pro vider -dependenc y and go v ernance limitations reduce their o v erall suitability in this case. The proposed frame w ork can support similar deplo yment-selection decisions in other constrained emer ging mark ets, pro vided that KPI weights and cost assumptions are recalibrated to the local re gulatory , traf c, cloud- readiness, and b usiness en vironment. A CKNO WLEDGMENTS The authors thank all W ataniya Mobile participants who contrib uted through intervie ws and techni cal discussions. This research w as not supported by an y e xternal research grant or contract. The article processing and publication char ges were self-funded by the authors. FUNDING INFORMA TION Authors state no funding in v olv ed. The w ork and publication char ges were self-funded by the authors. A UTHOR CONTRIB UTIONS ST A TEMENT This journal uses the Contri b ut or Roles T axonomy (CRediT) to recognize indi vidual author contrib u- tions, reduce authorship disputes, and f acilitate collaboration. Name of A uthor C M So V a F o I R D O E V i Su P Fu Saed T arapiah ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓ Shadi Atalla ✓ ✓ ✓ ✓ ✓ Kha wla Haz Daraghme ✓ ✓ ✓ ✓ ✓ C : C onceptualization I : I n v estig ation V i : V i sualization M : M ethodology R : R esources Su : Su pervision So : So ftw are D : D ata Curation P : P roject Administration V a : V a lidation O : Writing - O riginal Draft Fu : Fu nding Acquisition F o : F o rmal Analysis E : Writing - Re vie w & E diting CONFLICT OF INTEREST ST A TEMENT Authors state no conict of interest. V irtualized e volved pac k et cor e deployment str ate gies in emer ging P alestinian ... (Saed T ar apiah) Evaluation Warning : The document was created with Spire.PDF for Python.