Co esponding au ho : E geny Be eza
Copy igh © 2025 Au ho (s) e ain he copy igh o his a icle. This a icle is published unde he e ms o he C ea i e Commons A ibu ion Liscense 4.0.
Compa a i e analysis o mig a ion be ween cloud p o ide s (AWS and Google
Cloud) as a cos op imiza ion s a egy o high-load sys ems
E geny Be eza *
Head o So wa e, Ga ewise, Ma hash an, Is ael.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(01), 4225–4231
Publica ion his o y: Recei ed on 01 Ap il 2025; e ised on 07 Ap il 2025; accep ed on 14 Ap il 2025
A icle DOI: h ps://doi.o g/10.30574/wja .2025.26.1.1352
Abs ac
The s udy ocuses on conduc ing a compa a i e analysis o he mig a ion p ocess be ween wo leading cloud se ice
p o ide s: Amazon Web Se ices and Google Cloud, conside ed as means o cos op imiza ion o high-load sys ems.
The aim o he wo k is o iden i y key economic and echnical de e minan s de ining he o al cos o owne ship (TCO)
when changing cloud p o ide s, as well as o build a subs an ia ed decision-making model o ca ying ou such
mig a ion. As a me hodological basis, analysis o scien i ic a icles and indus y epo s o he pe iod 2021–2025,
compa ison o p icing models and pe o mance me ics o co e cloud se ices (compu e ins ances, da a s o age sys ems,
ne wo k componen s) a e used. The esul s o he s udy demons a e ha despi e signi ican ini ial in es men s in
mig a ion, including cos s o da a eg ess and a chi ec u al e inemen s, he s a egic ans e o wo kloads o he
pla o m wi h a mo e a o able p icing s uc u e (in pa icula , o Google Cloud wi h i s Sus ained Use Discoun s
p og am) is capable o ensu ing a educ ion in ope a ional expenses in he long e m. Based on he ob ained da a, a
decision-making ma ix is p oposed, sys ema izing he c i e ia o selec ing he a ge cloud pla o m depending on he
speci ics o he wo kload, expense p o ile and quali y-o -se ice equi emen s. The p esen ed conclusions and oolki
will be use ul o echnical di ec o s, heads o IT depa men s and cloud solu ion a chi ec s in s a egic planning and
op imiza ion o IT in as uc u e.
Keywo ds: Cloud Mig a ion; Cos Op imiza ion; High-Load Sys ems; AWS; Google Cloud; To al Cos O Owne ship;
Da a Eg ess; Mul i-Cloud S a egy; FinOps; Vendo Lock-In.
1. In oduc ion
In he con ex o he apid expansion o he olumes o gene a ed and p ocessed in o ma ion in he mode n digi al
economy, equi emen s o he scalabili y, aul ole ance and h oughpu capaci y o IT in as uc u es a e in ensi ying.
Cloud compu ing has es ablished i sel as he p ima y mechanism o deploying high-load sys ems, p o iding dynamic
scaling and pay-as-you-go p icing. Acco ding o es ima es by In e na ional Da a Co po a ion (IDC), global in es men s
in public cloud se ices will each USD 670 billion by he end o 2024 and exceed USD 1 illion by 2027, wi h a
compound annual g ow h a e (CAGR) o 19.9 % [7]. A he same ime, Amazon Web Se ices (AWS) and Mic oso
Azu e con inue o e ain leading posi ions, while Google Cloud Pla o m (GCP) demons a es ac i e sha e g ow h,
especially in he da a analy ics and machine lea ning segmen s [6].
The ele ance o he s udy is de e mined by he inc easing complexi y o he endo lock-in issue and he need o
con inuous imp o emen o ope a ional expendi u es (OpEx) in he ace o in ensi ying compe i ion. Acco ding o he
Flexe a S a e o he Cloud Repo o 2024, cos op imiza ion o exis ing cloud expendi u es is a p io i y o 89 % o
o ganiza ions [5]. A p o ide choice made se e al yea s ago may lose i s economic jus i ica ion o e ime due o he
e olu ion o applica ion a chi ec u es, changes in wo kload p o iles o he in oduc ion by compe i o s o mo e
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(01), 4225–4231
4226
ad an ageous p icing models. In his con ex , in e -cloud mig a ion ansi ions om a measu e o las eso o a sys em
o planned inancial managemen .
The scien i ic gap consis s in he insu icien codi ica ion and sys ema iza ion o mig a ion me hodologies speci ically
o high-load solu ions wi h a ocus on he compa a i e analysis o AWS and Google Cloud. Exis ing s udies a e mainly
limi ed ei he o a gene al o e iew o cloud se ices [2] o o p ocedu es o mig a ing om on-p emises in as uc u e
o he cloud [11], o e looking economic and echnical ba ie s o in e -cloud ansi ion, such as da a eg ess ees,
challenges in mig a ing managed da abases and he need o pe sonnel e aining.
The objec i e o he wo k is o iden i y he key economic and echnical de e minan s ha de ine he o al cos o
owne ship (TCO) when changing a cloud p o ide , as well as o cons uc a jus i ied decision-making model o
pe o ming such mig a ion.
The scien i ic no el y o he wo k mani es s in he sys ema iza ion o c i e ia o selec ing he a ge cloud pla o m and
he p oposal o a decision-making model o mig a ing high-load sys ems, in which he p o i abili y analysis
encompasses bo h di ec and indi ec cos s.
The au ho ’s hypo hesis is ha a s a egically calib a ed in e -cloud mig a ion, based on analysis o a chi ec u al
dependencies and p icing mechanisms, can educe long- e m ope a ional expenses despi e signi ican ini ial
in es men s in he mig a ion p ocess.
2. Ma e ials and me hods
In ecen yea s e iews o cloud pla o ms ha e paid special a en ion o compa ing he unc ionali y pe o mance and
p icing models o AWS and Google Cloud as a key ool o cos op imiza ion in high-load sys ems. Thus Gup a B., Mi al
P., Mu i T. [1] conduc a se ice-o ien ed analysis: he au ho s sys ema ize o e ings o compu e ins ances s o age and
ne wo king se ices and hen on he basis o p ice lis s cons uc a basic cos compa ison model. Kaushik P. e al. [2]
p o ile CPU load ope a ional I/O la encies and ne wo k me ics on equi alen AWS and GCP con igu a ions
demons a ing ha unde homogeneous wo kloads AWS deli e s highe CPU pe o mance whe eas GCP o e s
ad an ageous cold da a s o age p icing.
Au ho s Gohil R., Pa el H. [3] emphasize he con enience o cos -calcula ion ools and in eg a ion wi h CI/CD
in as uc u e: hey show how au oma ed p ice calcula o s enable apid modeling o mig a ion scena ios ye poin ou
signi ican a iabili y in inal igu es depending on egional a i s. Bo a P. [4] p oposes an adap i e p o ide -selec ion
s a egy: based on analysis o a la ge se o p icing plans he au ho demons a es ha an op imal mixed app oach may
in ol e placing ho wo kloads in AWS and long- e m a chi es in GCP.
Indus y epo s supplemen academic s udies wi h global s a is ics and iewpoin s o leading analy ical agencies.
Flexe a 2024 S a e o he Cloud Repo [5] eco ds he g owing popula i y o esou ce ese a ion s a egies and he
use o spo ins ances o subs an ial cos educ ion wi h 68 % o o ganiza ions planning o inc ease spending on FinOps
ools. Acco ding o Ga ne ® [6] AWS i mly main ains leade ship in pla o m b ead h and ecosys em ma u i y whe eas
GCP ea ns p aise o inno a ions in big da a and machine lea ning. IDC [7] con i ms ha he o al public cloud se ices
ma ke eached $669.2 billion in 2023 wi h GCP’s sha e showing he highes g ow h a e o app oxima ely 30 % yea -
o e -yea .
In he segmen o hyb id and mul i-cloud s a egies Anh N. H. [8] p oposes a classi ica ion o mig a ion app oaches (li
and shi epla o m e ac o ) compa ing hem by c i e ia o lexibili y secu i y and economic e iciency. Acco ding o
hei decision-making model o high-load sys ems he epla o m s age achie es an op imal balance be ween
a chi ec u e e ac o ing cos s and pe o mance gains. Me seedi K. J., Zeeba ee S. R. M. [9] desc ibe mechanisms o
o ches a ing da a and load be ween AWS and GCP o minimize ne wo k expenses and isks o endo lock-in.
Case s udies and economic analyses help o e eal he eal e ec s o mig a ion. In Heal h Ca e Cloud Mig a ion Case
S udy Deloi e US [11] he mig a ion o an EMR sys em o GCP is desc ibed whe e h ough au oma ic scaling and
edis ibu ion o idle esou ces i was possible o educe o al cos s by 30 %. Thallam N. S. T. [10] in u n pe o m a
comp ehensi e TCO analysis o ansi ioning HPC wo kloads om on-p emise o he cloud showing educ ion in cos s
while wa ning o hidden expenses o da a ans e and managemen se ices.
Thus despi e ela i e consis ency in compa ison me hodologies (se ice ca alogs plus syn he ic benchma ks) esul s
o en con adic each o he : some au ho s p io i ize AWS on he basis o aw pe o mance me ics [2] o he s a o GCP
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(01), 4225–4231
4227
o i s accommoda ing s o age and analy ics se ice p icing [4]. Indus y epo s do no always ake in o accoun he
speci ics o high-load sys ems and inhe en mig a ion o e heads ocusing ins ead on gene al ends [5, 6, 7]. Li e a u e
on hyb id s a egies clea ly desc ibes echnical and o ganiza ional models bu does no p o ide quan i a i e
assessmen s o he cos e sus esilience ade-o unde eal wo kloads [8, 9]. Meanwhile case s udies and inancial
analyses ocus on immedia e TCO o e looking long- e m e ec s such as de e io a ion o applica ion po abili y o
cul u al ba ie s wi hin o ganiza ions [10, 11]. The e o e u he in-dep h esea ch in o end- o-end owne ship cos
me ics o high-load sys ems ha conside s ull mig a ion li e cycles and mixed a chi ec u es is equi ed.
3. Resul s and Discussion
Based on he conduc ed analysis o exis ing esea ch, a me hodology o compa a i e analysis o high-load sys em
mig a ion p ocesses be ween AWS and Google Cloud pla o ms is p oposed, including he e alua ion o a chi ec u al
solu ions, inancial pe o mance modeling and he de elopmen o c i e ia o a jus i ied choice.
A ypical high-load in as uc u e comp ises a load balance o incoming a ic, a clus e o au oma ically scalable web
and/o applica ion se e s, a managed ela ional o NoSQL DBMS, a caching laye and objec s o age o s a ic con en .
Mig a ing such an en i onmen equi es no only a li -and-shi ans e o i ual ins ances bu also he adap a ion o
he a chi ec u e o he na i e se ices o he a ge cloud pla o m in o de o enhance e iciency and op imize cos s
(app oaches e ac o ing o e-a chi ec ing) [1, 2, 4].
Figu e 1 p esen s he schema o a high-load web applica ion adop ed as he baseline model o u he analysis.
Figu e 1 Typical a chi ec u e o a high-load web applica ion (compiled by he au ho based on [1, 2, 4]).
The compa ison o key AWS and GCP se ices ele an o his a chi ec u e is p esen ed in Table 1.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(01), 4225–4231
4228
Table 1 Compa ison o key AWS and Google Cloud se ices o high-load sys ems (compiled by he au ho based on [2,
3, 4, 10]).
Se ice
ca ego y
Amazon Web
Se ices (AWS)
Google Cloud
(GCP)
Key di e ences in he con ex o mig a ion and
cos s
Compu e
esou ces (VM)
Amazon EC2
(Elas ic Compu e
Cloud)
Google Compu e
Engine (GCE)
GCP o e s au oma ic sus ained use discoun s (SUDs)
which a e mo e bene icial o s able wo kloads
wi hou he need o up on paymen . AWS equi es
up on paymen (Sa ings Plans/RIs) o ob ain
maximum discoun s.
Objec s o age
Amazon S3 (Simple
S o age Se ice)
Google Cloud
S o age
P icing is compa able bu he cos o ope a ions GET
PUT and eg ess may di e . GCP o en o e s lowe
p ices o cold s o age classes.
Managed
ela ional
da abases
Amazon RDS
(Rela ional
Da abase Se ice)
Google Cloud SQL
Da abase mig a ion is he mos complex s age. GCP
Cloud SQL o e s simple in eg a ion wi h o he
Google se ices. Cos depends on pe o mance CPU
RAM IOPS.
Load balancing
Elas ic Load
Balancing
(ALB/NLB)
Cloud Load
Balancing
Bo h p o ide s o e global load balance s. GCP
p icing may be mo e p edic able as i includes da a
p ocessing and an hou ly a e pe o wa ding ule.
Ne wo k and
eg ess a ic
AWS Global
Accele a o / VPC
Google Cloud VPC /
CDN In e connec
A c i ical cos ac o . Eg ess cos s a AWS a e
adi ionally highe . GCP o e s a mo e gene ous ee
ie and lowe pe GB p ices a high olumes.
The p ocess o in e -cloud mig a ion can be di ided in o ou main s ages, as shown in Figu e 2.
Conduc ed analysis con i ms he ini ial hypo hesis: mig a ing high-load sys ems om AWS o Google Cloud can ensu e
a s able educ ion o cos s. A he same ime he decision o mig a e should no ely exclusi ely on compa ing egula
paymen s. The ollowing impo an ac o s come o he o e on :
Expenses o ou going a ic (Da a Eg ess). This is a one- ime bu signi ican cos i em ha mus be me iculously
modelled and calcula ed al eady a he p elimina y assessmen s age.
Complexi y o da abase ans e . The eloca ion o oluminous and esou ce-in ensi e eposi o ies is associa ed wi h a
high isk o down ime and po en ial da a loss, and he e o e equi es de elopmen o a de ailed mig a ion plan and
allback s a egies.
Human ac o . The p esence o compe encies in GCP wi hin he eam is a co ne s one condi ion o a success ul
ansi ion and s able ope a ion o he new in as uc u e [2, 4].
In he u he cou se o he esea ch ecommenda ions will be de eloped in de ail and subs an ia ed ega ding he
mig a ion o high-load sys ems be ween he AWS and Google Cloud pla o ms wi h he aim o cos op imiza ion.
To begin wi h i is ecommended o conduc a comp ehensi e audi o he cu en in as uc u e and wo kloads,
collec ing me ics on CPU usage, memo y, disk space and ne wo k a ic. Such an assessmen will allow o
iden i ica ion o bo lenecks and de e mina ion o po en ial a eas o sa ings when ans e ing esou ces be ween cloud
p o ide s. Simul aneously he o al cos o owne ship (TCO), including licensing, so wa e suppo and license ees,
should be assessed o co ec ly compa e cu en expenses wi h he o ecas ed cos s on AWS and Google Cloud.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(01), 4225–4231
4229
Figu e 2 S ages o he in e cloud mig a ion p ocess (compiled by he au ho based on [8, 9, 10]).
When analyzing he p icing models o cloud se ices i is ad isable o dis inguish he key cos d i e s: compu ing
esou ces, da a s o age, ou bound a ic and specialized se ices ( o example da abases o e en p ocessing sys ems).
I is ecommended o align he speci ica ions o AWS i ual machines (EC2, Fa ga e) wi h he analogous capabili ies o
Google Cloud (Compu e Engine, Cloud Run), aking in o accoun di e ences in billing o I/O and ne wo k ans e . This
app oach makes i possible o o ecas cos s in ad ance and o de e mine he op imal ins ance ypes o each ask.
The choice o mig a ion s a egy should be based on he ma u i y o he applica ions and he equi ed le el o
a chi ec u al modi ica ion. To minimize ini ial in es men s a li -and-shi app oach is o en employed wi h subsequen
p ope e ac o ing o cloud se ices. A he same ime i is ad isable o conside con aine iza ion ia Kube ne es (EKS
s GKE) o a se e less app oach (AWS Lambda s Google Cloud Func ions), which in he long e m will ensu e g ea e
scalabili y and educed cos s o idle esou ces.
As pa o mig a ion p epa a ion i is ecommended o es he ools o e ed by bo h p o ide s: AWS Mig a ion Hub and
Da abase Mig a ion Se ice, as well as Google Mig a e o Compu e Engine and Da abase Mig a ion Se ice. Conduc ing
a pilo p oo -o -concep on c i ically impo an subsys ems will allow alida ion o da a ans e scena ios, con igu a ion
o ne wo k policies and secu i y oles, he eby minimizing he isks o down ime and un o eseen expenses.
When designing he a ge a chi ec u e special a en ion should be paid o au oma ic scaling and in as uc u e as code
(In as uc u e as Code). The use o Te a o m o Cloud Deploymen Manage will enable cen alized esou ce
managemen in AWS and Google Cloud, uni ying deploymen p ocesses and simpli ying con igu a ion e iews. I is also
impo an o con igu e au o-scaling policies ha accoun o seasonal and peak loads, and o apply a clus e wa m-up
s a egy o ensu e eadiness o a ic g ow h.
Assessmen : Audi o cu en in as uc u e,
iden i ica ion o dependencies, p elimina y TCO
calcula ion on bo h pla o ms.
Planning: De elop a de ailed mig a ion plan, selec
ools (e.g. Google Cloud Mig a e o Compu e Engine),
design he a ge a chi ec u e on GCP, plan down ime
windows.
Execu ion: Mig a ing componen s in s ages: i s he
de elopmen and es ing en i onmen , hen he da a
mig a ion ( he mos expensi e s ep), and inally
swi ching DNS o he new in as uc u e.
Op imiza ion: Pos -mig a ion - se ing up au oscaling,
FinOps policies, moni o ing and secu i y on he new
pla o m.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(01), 4225–4231
4230
To main ain inancial discipline i is ecommended o implemen FinOps p ac ices: es ablish budge s and cos ale s,
implemen a esou ce agging sys em o de ail expenses by p ojec and eam, and in eg a e cloud APIs o au oma ic
da a expo o a BI sys em. This app oach ensu es anspa ency o expendi u es and enables apid iden i ica ion o
anomalies ela ed o cos inc eases.
A e mig a ion comple ion billing pa ame e s should be op imized: acqui e Rese ed Ins ances o Sa ings Plans in
AWS and Commi ed Use Discoun s in Google Cloud o long- e m, p edic able wo kloads. I is also ad isable o use spo
ins ances (AWS Spo Ins ances, Google P eemp ible VMs) o un egula ed o backg ound asks, which will p o ide
signi ican cos educ ion compa ed o on-demand esou ces.
Finally, i is ecommended o es ablish a p ocess o con inual imp o emen : egula ly e iew he cloud usage s a egy,
analyze he e ec i eness o implemen ed op imiza ions, and ain eams on new p o ide ea u es. Such a cul u e o
con inuous enhancemen will ensu e a chi ec u al adap abili y and maximize he e u n on in es men in cloud
in as uc u e.
4. Conclusion
In his s udy an analysis was conduc ed o he mig a ion o esou ce-in ensi e in o ma ion sys ems be ween he AWS
and Google Cloud pla o ms as a mechanism o cos op imiza ion. I was ound ha al hough AWS e ains a leading
ma ke posi ion, Google Cloud can o e compa able and in some cases mo e ad an ageous ope a ing condi ions unde
s able and p edic able wo kloads, p ima ily owing o i s long- e m discoun policy and lowe eg ess ne wo k a ic
a es.
The p incipal conclusion o his wo k is he con i ma ion ha mul i-cloud mig a ion cons i u es an e ec i e ye
inhe en ly complex ool o s a egic managemen o IT expendi u es. The p oposed me hodology comp ises a phased
mig a ion plan, a de ailed cos calcula ion model and a decision-making ma ix, enabling o maliza ion o he e alua ion
p ocess and minimiza ion o isks associa ed wi h he ansi ion. The esul s ob ained challenge he pe cep ion o
mig a ion as an exclusi ely echnical ask, posi ioning i ins ead as an in eg al elemen o co po a e inancial s a egy in
he ield o cloud echnologies (FinOps).
P ospec s o u he esea ch may in ol e adap ing he model o speci ic ypes o wo kloads, o example se e less
a chi ec u es o high-pe o mance compu ing (HPC), as well as de eloping me hods o educe indi ec cos s a ising
om empo a y pe o mance deg ada ion and he need o pe sonnel adap a ion.
Re e ences
[1] Gup a B., Mi al P., Mu i T. A e iew on amazon web se ice (aws), mic oso azu e & google cloud pla o m (gcp)
se ices //P oceedings o he 2nd In e na ional Con e ence on ICT o Digi al, Sma , and Sus ainable
De elopmen , ICIDSSD 2020, 27-28 Feb ua y 2020, Jamia Hamda d, New Delhi, India. – 2021. – pp. 1-9.
[2] Kaushik P. e al. Cloud compu ing and compa ison based on se ice and pe o mance be ween Amazon AWS,
Mic oso Azu e, and Google Cloud //2021 In e na ional Con e ence on Technological Ad ancemen s and
Inno a ions (ICTAI). – IEEE, 2021. – pp. 268-273.
[3] Gohil R., Pa el H. Compa a i e Analysis o Cloud Pla o m: Amazon Web Se ice, Mic oso Azu e, And Google
Cloud P o ide : A Re iew //2024 15 h In e na ional Con e ence on Compu ing Communica ion and Ne wo king
Technologies (ICCCNT). – IEEE, 2024. – pp. 1-5.
[4] Bo a P. Compa ison and analysis o leading cloud se ice p o ide s (AWS, Azu e and GCP) //In e na ional
Jou nal o Ad anced Resea ch in Enginee ing and Technology (IJARET) Volume. – 2024. – Vol. 15. – pp. 266-278.
h ps://doi.o g/10.17605/OSF.IO/T2DHW.
[5] Flexe a 2024 S a e o he Cloud Repo . [Elec onic esou ce]. – Access mode:
h ps://www.m3com a.com/whi e-pape s/ lexe a-s a e-o - he-cloud- epo -2024 (da e o access:
20.03.2025).
[6] Ga ne ® Magic Quad an ™ o S a egic Cloud Pla o m Se ices // Google Cloud. [Elec onic esou ce]. – Access
mode: h ps://cloud.google.com/ esou ces/con en /ga ne -s a egic-cloud-pla o m-se ices (accessed:
02/30/2025).
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(01), 4225–4231
4231
[7] IDC: Public cloud se ices ma ke g ows o $669.2 billion in 2023. [Elec onic esou ce]. – Access mode:
h ps://backendnews.ne /idc-public-cloud-se ices-ma ke -g ows- o-669-2-billion-in-2023/ (da e o access:
02/25/2025).
[8] Anh N. H. Hyb id Cloud Mig a ion S a egies: Balancing Flexibili y, Secu i y, and Cos in a Mul i-Cloud
En i onmen //T ansac ions on Machine Lea ning, A i icial In elligence, and Ad anced In elligen Sys ems. –
2024. – Vol. 14 (10). – pp. 14-26.
[9] Me seedi K. J., Zeeba ee S. R. M. The cloud a chi ec u es o dis ibu ed mul i-cloud compu ing: a e iew o hyb id
and ede a ed cloud en i onmen //The Indonesian Jou nal o Compu e Science. – 2024. – Vol. 13 (2). – pp.
1644-1673. h ps://doi.o g/10.33022/ijcs. 13i2.3811.
[10] Thallam N. S. T. Compa a i e Analysis o Public Cloud P o ide s o Big Da a Analy ics: AWS, Azu e, and Google
Cloud //In e na ional Jou nal o AI, BigDa a, Compu a ional and Managemen S udies. – 2023. – Vol. 4 (3). – pp.
18-29.
[11] Heal h Ca e Cloud Mig a ion Case S udy // Deloi e US. [Elec onic esou ce]. – Access mode:
h ps://www.deloi e.com/us/en/wha -we-do/capabili ies/cloud- ans o ma ion/case-s udies/heal h-ca e-
cloud-mig a ion-case-s udy.h ml (da e o access: 07.03.2025).
