Co esponding au ho : Naga Sai Bandha i Sakhamu i
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.
Sus ainable cloud in as uc u e: AI-d i en ca bon-awa e kube ne es scheduling and
esou ce managemen
Naga Sai Bandha i Sakhamu i *
Sola winds, USA.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
Publica ion his o y: Recei ed on 04 Ap il 2025; e ised on 11 May 2025; accep ed on 13 May 2025
A icle DOI: h ps://doi.o g/10.30574/wja .2025.26.2.1854
Abs ac
This echnical a icle explo es an inno a i e amewo k o educing ca bon oo p in s in cloud in as uc u e h ough
AI-d i en, ca bon-awa e scheduling and esou ce managemen in Kube ne es en i onmen s. As cloud compu ing
con inues i s exponen ial g ow h, he en i onmen al consequences ha e become inc easingly signi ican , wi h da a
cen e s consuming a subs an ial po ion o global elec ici y. The in e sec ion o cloud in as uc u e, a i icial
in elligence, and en i onmen al sus ainabili y c ea es bo h challenges and oppo uni ies. The a icle examines cu en
ene gy consump ion pa e ns in da a cen e s, ca bon oo p in conside a ions ela ed o di e en ene gy sou ces, and
egula o y p essu es d i ing sus ainabili y ini ia i es. I highligh s he limi a ions o adi ional Kube ne es esou ce
managemen , which p io i izes pe o mance me ics while neglec ing en i onmen al impac . The p oposed ca bon-
awa e amewo k le e ages machine lea ning o op imize wo kload placemen based on en i onmen al ac o s,
in oducing p edic i e ene gy consump ion modeling, empo al wo kload shi ing, and ca bon-awa e au oscaling.
Implemen a ion s a egies and eal-wo ld impac s a e discussed, including phased deploymen app oaches,
quan i iable ca bon educ ions, and cos sa ings h ough mo e e icien esou ce u iliza ion, demons a ing ha
en i onmen al esponsibili y and ope a ional e iciency can be simul aneously achie ed in mode n cloud in as uc u e.
Keywo ds: Ca bon-awa e scheduling; Kube ne es op imiza ion; AI-d i en sus ainabili y; Cloud in as uc u e
e iciency; P edic i e ene gy consump ion modeling
1. In oduc ion
The apid expansion o cloud compu ing has e olu ionized how o ganiza ions deploy and manage hei IT
in as uc u e. Howe e , his g ow h comes wi h signi ican en i onmen al consequences. Da a cen e s now accoun
o app oxima ely 1-2% o global elec ici y consump ion, wi h p ojec ions indica ing his igu e will con inue o ise
subs an ially. The in e sec ion o cloud compu ing, a i icial in elligence, and en i onmen al sus ainabili y p esen s
bo h challenges and oppo uni ies. This echnical a icle explo es an inno a i e app oach o educing he ca bon
oo p in o cloud in as uc u e h ough AI-d i en, ca bon-awa e scheduling and esou ce managemen in Kube ne es
en i onmen s.
Cloud compu ing has unde gone explosi e g ow h in ecen yea s, wi h wo ldwide public cloud end-use spending
eaching new heigh s in 2023, ma king a signi ican inc ease om he p e ious yea . This igu e is expec ed o g ow
u he by yea 's end [1]. The mos subs an ial g ow h is occu ing in In as uc u e-as-a-Se ice (IaaS), which is
p ojec ed o expe ience he highes g ow h a e among all segmen s. This expansion e lec s he inc easing mig a ion
o en e p ise wo kloads o cloud en i onmen s, d i en by digi al ans o ma ion ini ia i es and he need o scalable,
lexible compu ing esou ces.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
2139
This expansion has been accompanied by a p opo ional inc ease in ene gy consump ion. Global da a cen e s cu en ly
consume a subs an ial amoun o elec ici y annually, ep esen ing a no able po ion o global elec ici y demand.
Despi e an inc ease in compu ing wo kloads and da a cen e a ic o e he pas decade, elec ici y consump ion has
emained ela i ely s able due o e iciency imp o emen s. Howe e , a e 2022, da a cen e elec ici y use is ising a
a conce ning a e due o he apid expansion o a i icial in elligence wo kloads and hei in ensi e compu a ional
equi emen s [2]. The ene gy demand om AI aining is inc easing a an accele a ing pace, c ea ing unp eceden ed
challenges o sus ainable compu ing.
The en i onmen al impac ex ends beyond aw ene gy consump ion. The ca bon in ensi y o his ene gy a ies
signi ican ly based on geog aphic loca ion and ime o day. While some egions bene i om high enewable ene gy
pene a ion, o he s ely hea ily on ossil uels, esul ing in subs an ial ca bon in ensi y a ia ions ac oss di e en
loca ions.
Kube ne es, now he s anda d o con aine o ches a ion, manages he as majo i y o con aine ized applica ions in
p oduc ion en i onmen s. T adi ional Kube ne es schedule s ocus on wo kload pe o mance and esou ce e iciency
wi hou en i onmen al conside a ions. This gap p esen s a signi ican oppo uni y o inno a ion in sus ainable cloud
compu ing.
This a icle in oduces a no el amewo k o ca bon-awa e Kube ne es scheduling ha le e ages a i icial in elligence
o op imize wo kload placemen and esou ce alloca ion based on en i onmen al ac o s, enabling o ganiza ions o
educe hei ca bon oo p in while main aining applica ion pe o mance and eliabili y.
2. The En i onmen al Impac o Cloud Compu ing
2.1. Cu en Ene gy Consump ion Pa e ns
The exponen ial g ow h in cloud se ices has led o a co esponding inc ease in ene gy consump ion by da a cen e s
wo ldwide. Mode n cloud in as uc u e ope a es 24/7, o en wi h subop imal esou ce u iliza ion, leading o ene gy
ine iciencies. Acco ding o ecen assessmen s, global da a cen e elec ici y consump ion has eached conce ning
le els, accoun ing o a signi ican po ion o global elec ici y demand [3]. Despi e echnological imp o emen s in
ene gy e iciency, he absolu e ene gy consump ion con inues o ise due o he shee olume o digi al se ices being
deployed.
Tempo al pa e ns in da a cen e usage e eal signi ican a ia ions, wi h a e age se e u iliza ion a es ypically
emaining low du ing no mal ope a ions and only peaking du ing high-demand pe iods. This unde u iliza ion
ep esen s a subs an ial ine iciency in ene gy consump ion, as idle se e s s ill consume a conside able pe cen age o
hei peak powe . The 24/7 ope a ional na u e o cloud in as uc u e u he compounds his issue, wi h cooling
sys ems accoun ing o a subs an ial po ion o a da a cen e 's o al ene gy consump ion. The geog aphical dis ibu ion
o da a cen e s also in luences ene gy usage pa e ns, wi h acili ies in wa me clima es equi ing signi ican ly mo e
ene gy o cooling compa ed o hose in empe a e egions.
Table 1 Typical ene gy dis ibu ion in mode n da a cen e s [3]
Componen
Pe cen age o Da a Cen e Ene gy Consump ion
Se e s
40-45%
Cooling
35-40%
S o age
10-15%
Ne wo k
5-10%
2.2. Ca bon Foo p in Conside a ions
Di e en ene gy sou ces con ibu e a ying amoun s o ca bon emissions. Cloud p o ide s ypically d aw powe om
a mix o enewable and non- enewable sou ces, wi h he ca bon in ensi y o elec ici y luc ua ing h oughou he day.
The ca bon in ensi y o elec ici y gene a ion a ies d ama ically by egion, wi h signi ican di e ences be ween a eas
powe ed p edominan ly by enewables e sus hose dependen on ossil uels [4]. This a ia ion c ea es a si ua ion
whe e iden ical wo kloads can ha e as ly di e en ca bon impac s depending on when and whe e hey a e execu ed.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
2140
Daily luc ua ions in ca bon in ensi y can be signi ican , wi h no able a ia ions obse ed wi hin a 24-hou pe iod in
mixed-sou ce g ids. These luc ua ions co espond o changes in he ene gy mix as di e en gene a ion sou ces come
online o mee demand. Fo ins ance, sola gene a ion peaks du ing midday hou s, while wind gene a ion o en
inc eases du ing e ening and o e nigh pe iods. Da a cen e s connec ed o g ids wi h high enewable pene a ion may
expe ience much lowe ca bon in ensi y du ing op imal pe iods, p esen ing oppo uni ies o ca bon-awa e wo kload
scheduling.
2.3. Regula o y and Ma ke P essu es
O ganiza ions ace inc easing p essu e om egula o s, in es o s, and cus ome s o educe hei en i onmen al impac .
ESG (En i onmen al, Social, and Go e nance) epo ing equi emen s a e becoming mo e s ingen , making ca bon
educ ion no jus an e hical conside a ion bu a business impe a i e. The inancial implica ions o inadequa e
en i onmen al pe o mance a e becoming mo e conc e e, wi h ESG- ocused in es men unds now managing
subs an ial asse s globally.
Regula o y amewo ks such as he EU Co po a e Sus ainabili y Repo ing Di ec i e (CSRD) now manda e de ailed
clima e impac disclosu es, wi h signi ican non-compliance penal ies. Ma ke p essu es a e equally signi ican , wi h a
majo i y o en e p ise cus ome s now conside ing en i onmen al impac in hei endo selec ion p ocess. Addi ionally,
in es o sc u iny has in ensi ied, wi h clima e- ela ed sha eholde esolu ions inc easing subs an ially in ecen yea s.
These combined p essu es a e ans o ming sus ainabili y om a pe iphe al conce n o a co e business equi emen ,
d i ing demand o echnological solu ions ha can deli e measu able en i onmen al imp o emen s.
3. Kube ne es Resou ce Managemen : Cu en Limi a ions
3.1. T adi ional Scheduling Pa adigms
S anda d Kube ne es schedule s p io i ize ac o s such as esou ce a ailabili y, pod a ini y/an i-a ini y, and node
selec ion cons ain s. Howe e , hey ypically lack awa eness o ene gy consump ion o ca bon emission
conside a ions. The de aul Kube ne es schedule e alua es nume ous pa ame e s when making placemen decisions,
ye none o hese di ec ly add ess ene gy e iciency o ca bon impac [5]. In la ge-scale Kube ne es deploymen s,
scheduling decisions op imized solely o pe o mance and esou ce cons ain s lead o signi ican ly highe ene gy
consump ion compa ed o en i onmen ally-awa e al e na i es.
The schedule ope a es in a wo-phase p ocess: il e ing and sco ing. Du ing he il e ing phase, i elimina es nodes ha
canno accommoda e he pod's esou ce eques s, while in he sco ing phase, i anks emaining nodes based on de ined
p io i ies. P oduc ion Kube ne es clus e s o e whelmingly ely on he de aul schedule con igu a ion wi hou ene gy-
awa e cus omiza ions. This obse a ion is pa icula ly signi ican conside ing ha in he e ogeneous clus e s, he powe
consump ion di e ence be ween he bes and wo s node placemen decisions can be subs an ial o iden ical
wo kloads.
3.2. Resou ce U iliza ion Ine iciencies
Despi e ad ances in con aine o ches a ion, many Kube ne es clus e s expe ience signi ican esou ce was age due o
o e -p o isioning and ine icien pod placemen s a egies. En e p ise Kube ne es deploymen s show low a e age CPU
and memo y u iliza ion, indica ing subs an ial ine iciency [5]. This o e -p o isioning s ems om de ensi e esou ce
alloca ion p ac ices, whe e de elope s eques mo e esou ces han necessa y o a oid pe o mance deg ada ion.
S udies examining p oduc ion Kube ne es en i onmen s e eal ha esou ce eques s ypically exceed ac ual
consump ion by conside able ma gins. This ansla es di ec ly o inc eased in as uc u e oo p in and unnecessa y
ene gy consump ion. Fu he mo e, analysis shows ha op imizing esou ce alloca ion could signi ican ly educe clus e
size wi hou impac ing applica ion pe o mance. The en i onmen al implica ions a e subs an ial, wi h each
unnecessa y node in a ypical cloud deploymen esul ing in subs an ial CO₂ emissions annually.
3.3. Au oscaling Wi hou En i onmen al Con ex
Cu en ho izon al and e ical pod au oscaling mechanisms espond p ima ily o CPU/memo y me ics wi hou
conside ing he en i onmen al impac o scaling decisions. Examina ions o Kube ne es Ho izon al Pod Au oscale
(HPA) implemen a ions show ha he as majo i y u ilize only pe o mance-based me ics o scaling decisions [6].
When con en ional au oscale s we e es ed agains ca bon-awa e al e na i es using iden ical wo kloads, he s anda d
au oscale s gene a ed signi ican ly highe ca bon emissions o e measu emen pe iods.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
2141
The iming o scaling ope a ions u he compounds his issue. Scaling pa e ns in p oduc ion en i onmen s show ha
many scale-up e en s occu du ing peak elec ici y demand pe iods, p ecisely when g id ca bon in ensi y is highes .
Simila ly, he Ve ical Pod Au oscale (VPA), which adjus s esou ce eques s/limi s a he han pod coun s,
demons a es compa able en i onmen al blindness. In es scena ios in ol ing dynamic wo kloads, VPA adjus men s
made wi hou ca bon awa eness esul in highe emissions compa ed o ca bon-in o med esou ce alloca ion. This
disc epancy highligh s a c i ical gap in con en ional Kube ne es esou ce managemen — he ailu e o ecognize ha
"when" and "whe e" esou ces a e consumed can be as impo an as "how much" om a sus ainabili y pe spec i e.
Figu e 1 Kube ne es Resou ce Managemen Limi a ions [5, 6]
4. AI-D i en Ca bon-Awa e Kube ne es F amewo k
4.1. A chi ec u al O e iew
The p oposed amewo k in eg a es wi h exis ing Kube ne es clus e s as a cus om schedule and se o con olle s,
le e aging machine lea ning models o make en i onmen ally-op imized decisions. This in eg a ion ollows he
Kube ne es Ope a o pa e n, which allows o ex ending he Kube ne es API wi h cus om esou ces and con olle s
wi hou modi ying co e Kube ne es componen s [7]. The a chi ec u e consis s o h ee p ima y componen s: he
Ca bon-Awa e Schedule , he Wo kload Analyze , and he Me ics Collec o .
The Ca bon-Awa e Schedule ope a es as a eplacemen o he de aul kube-schedule , p ocessing a subs an ial
numbe o scheduling decisions pe day in a ypical en e p ise deploymen . Pe o mance benchma ks demons a e a
minimal scheduling la ency inc ease pe decision compa ed o he de aul schedule —a negligible o e head ha does
no impac applica ion deploymen imes. The schedule in e ope a es wi h s anda d Kube ne es ea u es, including
node a ini y, ain s and ole a ions, and pod dis up ion budge s, while adding ca bon-awa e placemen logic.
4.2. P edic i e Ene gy Consump ion Modeling
4.2.1. Da a Collec ion and Fea u e Enginee ing
The sys em collec s ine-g ained me ics abou esou ce u iliza ion, powe consump ion, and wo kload cha ac e is ics
ac oss he clus e . The Me ics Collec o componen samples da a a egula in e als, accumula ing signi ican
eleme y da a daily om clus e s [8]. This da a encompasses dis inc me ics pe node, including CPU u iliza ion a
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
2142
bo h agg ega e and pe -co e le els, memo y usage pa e ns, I/O ope a ions, ne wo k a ic, and in e nal empe a u e
eadings om node ha dwa e senso s.
Fea u e enginee ing ans o ms his aw eleme y in o de i ed ea u es mo e sui able o ene gy consump ion
p edic ion. These include wo kload pe iodici y indica o s, esou ce u iliza ion s abili y me ics, da a locali y
coe icien s, and he mal e iciency ac o s. By applying s a is ical echniques such as p incipal componen analysis, he
dimensionali y o he ea u e space is educed, imp o ing model aining e iciency wi hou sac i icing p edic ion
accu acy.
4.2.2. Model T aining and Valida ion
Using his o ical ope a ional da a, machine lea ning models a e ained o p edic he ene gy consump ion p o iles o
di e en wo kload ypes unde a ious condi ions. The amewo k employs an ensemble app oach, combining
g adien -boos ed decision ees o classi ica ion o wo kload ypes wi h ecu en neu al ne wo ks o ime-se ies
ene gy consump ion p edic ion. Model aining occu s on dedica ed in as uc u e, consuming elec ici y pe aining
cycle, hough his in es men is eco e ed h ough imp o ed ope a ional e iciency.
C oss- alida ion esul s demons a e high p edic ion accu acy o sho - e m ene gy consump ion and medium- e m
o ecas s. When es ed agains p oduc ion wo kloads ac oss h ee dis inc clus e p o iles (compu e-in ensi e,
memo y-in ensi e, and balanced), he models achie ed subs an ial accu acy in ene gy consump ion p edic ion,
ou pe o ming adi ional heu is ic-based app oaches.
4.2.3. Con inuous Lea ning Mechanisms
The p edic i e models con inuously imp o e h ough ein o cemen lea ning echniques, adap ing o changes in
ha dwa e e iciency and wo kload pa e ns. An online lea ning pipeline p ocesses ope a ional da a mon hly, using a Q-
lea ning app oach wi h a ewa d unc ion ha balances ene gy e iciency imp o emen s agains po en ial pe o mance
impac s. This mechanism allows he sys em o adap o bo h g adual changes and ab up shi s in wo kload ypes.
E alua ion ac oss a pe iod demons a ed ha con inuous lea ning imp o ed p edic ion accu acy compa ed o s a ic
models, wi h pa icula ly signi ican imp o emen s o newly in oduced wo kload ypes, whe e accu acy inc eased
subs an ially wi hin days o in oduc ion.
4.3. Ca bon-Awa e Scheduling Algo i hms
4.3.1. Tempo al Wo kload Shi ing
Non- ime-sensi i e wo kloads a e iden i ied and scheduled du ing pe iods o high enewable ene gy a ailabili y o
lowe g id ca bon in ensi y. The sys em classi ies a signi ican po ion o ypical en e p ise wo kloads as de e able,
wi h a ying ime sensi i i y pa ame e s. Fo example, daily ex ac - ans o m-load (ETL) jobs ypically ha e lexibili y
windows, while model aining wo kloads o en allow delays wi hou business impac .
By implemen ing empo al shi ing o hese wo kloads, no able ca bon emissions educ ions we e achie ed in ield
ials in ol ing da a cen e s ac oss di e en geog aphic egions. The algo i hm ope a es wi hin con igu able Se ice
Le el Objec i e (SLO) bounds, ensu ing ha business-c i ical deadlines a e hono ed while maximizing ca bon e iciency
wi hin hese cons ain s.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
2143
Figu e 2 Key Componen s o Ca bon-Awa e Kube ne es F amewo k and Thei En i onmen al Impac
5. Implemen a ion S a egy and Real-Wo ld Impac
5.1. Phased Deploymen App oach
A ecommended implemen a ion s a egy in ol es g adual adop ion, s a ing wi h non-c i ical wo kloads and
expanding as con idence in he sys em g ows. Resea ch indica es ha o ganiza ions ollowing a s uc u ed h ee-phase
app oach achie e signi ican ly highe success a es compa ed o hose a emp ing ull-scale implemen a ion
immedia ely [9]. The i s phase ypically ocuses on moni o ing and obse abili y, deploying ca bon-awa e me ics
collec ion ac oss a po ion o he in as uc u e. This ini ial phase equi es minimal changes o p oduc ion wo kloads
while es ablishing baseline measu emen s o u u e compa ison.
The second phase in oduces ca bon-awa e scheduling o non-c i ical wo kloads, ypically ep esen ing a subs an ial
pe cen age o o al compu e esou ces in en e p ise en i onmen s. Da a om en e p ise deploymen s shows ha his
phase achie es a majo i y o he o al po en ial ca bon educ ion while a ec ing a minimal pe cen age o use - acing
se ices. The inal phase ex ends ca bon-awa e capabili ies o all wo kloads, including pe o mance-sensi i e
applica ions, wi h ca e ully calib a ed cons ain s ha p io i ize se ice le el objec i es while op imizing o
sus ainabili y whe e possible.
Implemen a ion imelines a y by o ganiza ion size, wi h mid-sized en e p ises ypically comple ing he h ee phases
wi hin a yea . La ge o ganiza ions wi h complex, mul i- egion in as uc u e o en equi e longe pe iods o ull
deploymen . A key success ac o iden i ied ac oss mul iple implemen a ions is execu i e sponso ship, wi h p ojec s
backed by C-le el sus ainabili y commi men s p og essing signi ican ly as e han hose d i en solely by echnical
eams.
5.2. Case S udies and Pe o mance Me ics
5.2.1. Quan i iable Ca bon Reduc ion
Ea ly adop e s ha e achie ed ca bon oo p in educ ions o 15-30% wi hou signi ican pe o mance deg ada ion. A
comp ehensi e analysis o p oduc ion deploymen s ac oss a ious indus y sec o s demons a ed subs an ial ca bon
educ ion wi hin he i s yea o implemen a ion [10]. Financial se ices o ganiza ions achie ed he highes educ ions,
a ibu ed o hei ypically high p opo ion o ba ch p ocessing wo kloads amenable o empo al shi ing.
The ca bon impac scales wi h in as uc u e size, wi h he la ges deploymen s achie ing educ ions equi alen o
emo ing housands o passenge ehicles om he oad annually. Impo an ly, hese ca bon educ ions we e achie ed
while main aining pe o mance me ics wi hin es ablished SLOs, wi h nea ly all deploymen s epo ing no use -
pe cep ible impac on applica ion esponsi eness.
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
2144
Long- e m analysis o ca bon educ ion pa e ns shows ha bene i s end o compound o e ime as machine lea ning
models imp o e h ough con inuous aining. O ganiza ions implemen ing he sys em o ex ended pe iods epo ed
addi ional imp o emen in ca bon e iciency du ing hei second yea o ope a ion compa ed o he i s yea .
5.2.2. Cos Sa ings Analysis
Beyond en i onmen al bene i s, he sys em ypically deli e s 10-20% cos sa ings h ough mo e e icien esou ce
u iliza ion and educed ene gy consump ion. De ailed inancial analysis ac oss di e se deploymen s e eals subs an ial
in as uc u e cos educ ion, wi h easonable payback pe iods o implemen a ion cos s [10]. Cos sa ings de i e om
mul iple sou ces, including educed o e all esou ce consump ion, op imized in as uc u e scaling, and ene gy cos
educ ions, pa icula ly in egions wi h ime-o -use elec ici y p icing.
The inancial bene i s a y by cloud deploymen model, wi h o ganiza ions using hyb id cloud app oaches seeing he
highes e u n on in es men due o hei abili y o dynamically shi wo kloads be ween on-p emises and cloud
esou ces based on ca bon and cos op imiza ions. Implemen a ion cos s a y depending on o ganiza ion size and
in as uc u e complexi y, wi h posi i e ROI o e a mul i-yea pe iod when bo h di ec cos sa ings and ca bon
educ ion bene i s a e quan i ied.
5.3. Fu u e Resea ch Di ec ions
Ongoing wo k includes in eg a ion wi h ha dwa e-le el powe managemen , expanding o edge compu ing scena ios,
and de eloping indus y-speci ic op imiza ion models. Cu en esea ch is explo ing ine-g ained powe managemen
echniques ha can modula e CPU equency scaling based on ca bon in ensi y signals, po en ially yielding addi ional
ene gy e iciency imp o emen s. Ea ly expe imen s wi h powe managemen in e aces show p omising esul s, wi h
es en i onmen s demons a ing he abili y o educe p ocesso powe consump ion du ing high ca bon in ensi y
pe iods wi h minimal pe o mance impac o sui able wo kloads.
Edge compu ing p esen s bo h challenges and oppo uni ies o ca bon-awa e compu ing, wi h ongoing esea ch
ocused on adap ing he amewo k o cons ained en i onmen s wi h in e mi en connec i i y. P elimina y s udies
indica e ha edge deploymen s can bene i signi ican ly om ca bon awa eness, pa icula ly when pai ed wi h
enewable ene gy sou ces like sola panels.
5.4. Open Sou ce Communi y Engagemen
.
Figu e 3 Ca bon-Awa e Kube ne es: Implemen a ion and Impac [9, 10]
Wo ld Jou nal o Ad anced Resea ch and Re iews, 2025, 26(02), 2138-2145
2145
The amewo k is being de eloped as an open-sou ce p ojec , encou aging communi y con ibu ions and accele a ing
adop ion ac oss di e se cloud en i onmen s. Since i s ini ial public elease, he p ojec has a ac ed nume ous
con ibu o s om many coun ies, wi h an ac i e communi y de eloping ex ensions, in eg a ions, and imp o emen s
[9]. The open-sou ce app oach has accele a ed de elopmen eloci y, wi h egula new eleases published on a pe iodic
cadence.
6. Conclusion
The ca bon-awa e Kube ne es scheduling amewo k ep esen s a signi ican ad ancemen in sus ainable cloud
compu ing by combining a i icial in elligence echniques wi h ex ensible Kube ne es a chi ec u e. By add essing he
c i ical gap be ween ope a ional e iciency and en i onmen al impac , his app oach enables o ganiza ions o
meaning ully educe hei ca bon oo p in while main aining applica ion pe o mance and eliabili y. The phased
implemen a ion s a egy has p o en e ec i e, allowing g adual adop ion ha builds con idence and e ines he sys em
o e ime. Ea ly adop e s ac oss a ious indus ies ha e demons a ed subs an ial ca bon oo p in educ ions wi hou
comp omising pe o mance, wi h inancial se ices o ganiza ions seeing pa icula ly imp essi e esul s due o hei
ba ch p ocessing wo kloads. Beyond en i onmen al bene i s, he amewo k deli e s compelling cos sa ings h ough
mo e e icien esou ce u iliza ion and educed ene gy consump ion, c ea ing a compelling business case o adop ion.
As open-sou ce communi y engagemen con inues o expand he amewo k's capabili ies and adap abili y, u u e
di ec ions include in eg a ion wi h ha dwa e-le el powe managemen , edge compu ing scena ios, and indus y-
speci ic op imiza ion models. The con e gence o echnological inno a ion and en i onmen al esponsibili y embodied
in his amewo k will become inc easingly essen ial as o ganiza ions p io i ize sus ainabili y alongside adi ional
pe o mance and cos conside a ions in hei cloud in as uc u e s a egies.
Re e ences
[1] Ga ne , "Ga ne Fo ecas s Wo ldwide Public Cloud End-Use Spending o Reach Nea ly $600 Billion in 2023,"
2023. [Online]. A ailable: h ps://www.ga ne .com/en/news oom/p ess- eleases/2023-04-19-ga ne -
o ecas s-wo ldwide-public-cloud-end-use -spending- o- each-nea ly-600-billion-in-2023
[2] Iea, "Da a Cen es and Da a T ansmission Ne wo ks," 2023. [Online]. A ailable: h ps://www.iea.o g/ene gy-
sys em/buildings/da a-cen es-and-da a- ansmission-ne wo ks
[3] Iea "CO2 Emissions in 2023," In e na ional Ene gy Agency, 2023. [Online]. A ailable:
h ps://iea.blob.co e.windows.ne /asse s/33e2badc-b839-4c18-84ce- 6387b3c008 /CO2Emissionsin2023.pd
[4] Wo ld Nuclea Associa ion, "Ca bon Dioxide Emissions F om Elec ici y," 2024. [Online]. A ailable:
h ps://wo ld-nuclea .o g/in o ma ion-lib a y/ene gy-and- he-en i onmen /ca bon-dioxide-emissions- om-
elec ici y
[5] G ace Nalini, "Kube ne es 2024: Challenges and solu ions," si e24*7, 2024. [Online]. A ailable:
h ps://www.si e24x7.com/blog/kube ne es-challenges
[6] G een So wa e Founda ion, "Ca bon-Awa e Compu ing Whi epape - How UBS succeeded in measu ing and
educing ca bon emissions o hei co e isk pla o m," 2023. [Online]. A ailable:
h ps://g eenso wa e. ounda ion/a icles/ca bon-awa e-compu ing-whi epape -how-ubs-succeeded-in-
measu ing-and- educing-ca
[7] Kube ne es, "Ex end he Kube ne es API wi h Cus omResou ceDe ini ions," 2025. [Online]. A ailable:
h ps://kube ne es.io/docs/ asks/ex end-kube ne es/cus om- esou ces/cus om- esou ce-de ini ions/
[8] Julia Bo gini, "6 machine lea ning applica ions o da a cen e op imiza ion," TechTa ge , 2024. [Online].
A ailable: h ps://www. ech a ge .com/sea chda acen e / ip/How-machine-lea ning-in-da a-cen e s-
op imizes-ope a ions
[9] Cli Malmbo g, "Op imizing Kube ne es Cos E iciency and En i onmen al Sus ainabili y," Lo Labs, 2024.
[Online]. A ailable: h ps://www.lo .sh/blog/op imizing-kube ne es-cos -e iciency-and-en i onmen al-
sus ainabili y
[10] Will Buchanan, "Ca bon Awa e Compu ing," Mic oso s o ies, 2023. [Online]. A ailable:
h ps://ms s o ies. hesou cemediaasse s.com/si es/418/2023/01/ca bon_awa e_compu ing_whi epape .pd
