Submitted by: Binaya Kumar Padhi Regd. No.: 1302001377
BINAYA K!AR PA"#I
1302001377
Abstract
$%oud &om'uting i( o))ering uti%ity oriented I* (er+i&e( to u(er( ,or%d,ide. It enab%e( ho(ting o) a''%i&ation( )rom &on(umer- (&ienti)i& and bu(ine(( domain(. #o,e+er data &enter( ho(ting &%oud &om'uting a''%i&ation( &on(ume huge amount( o) energy- &ontributing to high o'erationa% &o(t( and &arbon )oot'rint( to the en+ironment. ith energy (hortage( and g%oba% &%imate &hange %eading our &on&ern( the(e day(- the 'o,er &on(um'tion o) data &enter( ha( be&ome a /ey i((ue. *here)ore- ,e need green &%oud &om'uting (o%ution( that &an not on%y (a+e energy- but a%(o redu&e o'erationa% &o(t(. *he +i(ion )or energy e))i&ient management o) &%oud &om'uting en+ironment( i( 're(ented here. A green (&hedu%ing a%gorithm ,hi&h ,or/( by 'o,ering do,n (er+er( ,hen they are not in u(e i( a%(o 're(ented. Green Cloud computing i( en+i(ioned to a&hie+e not on%y e))i&ient 'ro&e((ing and
uti%i(ation o) &om'uting in)ra(tru&ture- but a%(o minimi(e energy &on(um'tion. *hi( i( e((entia% )or en(uring that the )uture gro,th o) $%oud &om'uting i( (u(tainab%e. ther,i(e- $%oud &om'uting ,ith in&rea(ing%y 'er+a(i+e )rontend &%ient de+i&e( intera&ting ,ith ba&/end data &entre( ,i%% &au(e an enormou( e(&a%ation o) energy u(age.
1
BINAYA K!AR PA"#I
1302001377
Introduction
In 1- 4eonard K%einro&/ - one o) the &hie) (&ienti(t( o) the origina% Ad+an&ed Re(ear&h Pro5e&t( Agen&y Net,or/ 6ARPAN*8 ,hi&h (eeded the Internet- (aid: “As of now, computer networks are still in their infancy, but as they grow up and become sophisticated, we will probably see the spread of „computer utilities
‟
which, like present electric and telephone utilities, will
service individual homes and offices across the country.” *hi( +i(ion o) &om'uting uti%itie( ba(ed
on a (er+i&e 'ro+i(ioning mode% anti&i'ated the ma((i+e tran()ormation o) the entire &om'uting indu(try in the 21(t &entury ,hereby &om'uting (er+i&e( ,i%% be readi%y a+ai%ab%e on demand- %i/e other uti%ity (er+i&e( a+ai%ab%e in today9( (o&iety. Simi%ar%y- u(er( 6&on(umer(8 need to 'ay 'ro+ider( on%y ,hen they a&&e(( the &om'uting (er+i&e(. In addition- &on(umer( no %onger need to in+e(t hea+i%y or en&ounter di))i&u%tie( in bui%ding and maintaining &om'%e I* in)ra(tru&ture. In (u&h a mode%- u(er( a&&e(( (er+i&e( ba(ed on their re;uirement( ,ithout regard to ,here the (er+i&e( are ho(ted. *hi( mode% ha( been re)erred to a( utility computing - or re&ent%y a( Cloud computing . *he %atter term denote( the in)ra(tru&ture a( a <$%oud= )rom ,hi&h bu(ine((e( and
u(er( &an a&&e(( a''%i&ation( a( (er+i&e( )rom any,here in the ,or%d on demand. #en&e- $%oud &om'uting &an be &%a((i)ied a( a ne, 'aradigm )or the dynami& 'ro+i(ioning o) &om'uting (er+i&e( (u''orted by (tateo)theart data &enter( that u(ua%%y em'%oy >irtua% !a&hine 6>!8 te&hno%ogie( )or &on(o%idation and en+ironment i(o%ation 'ur'o(e( . !any &om'uting (er+i&e 'ro+ider( in&%uding ?oog%e- !i&ro(o)t- Yahoo- and IB! are ra'id%y de'%oying data &enter( in +ariou( %o&ation( around the ,or%d to de%i+er $%oud &om'uting (er+i&e(. $%oud &om'uting de%i+er( in)ra(tru&ture- '%at)orm- and (o)t,are 6a''%i&ation(8 a( (er+i&e(- ,hi&h are made a+ai%ab%e to &on(umer( a( (ub(&ri'tionba(ed (er+i&e( under the 'aya(yougo mode%. In indu(try the(e (er+i&e( are re)erred to a( In)ra(tru&ture a( a Ser+i&e 6IaaS8- P%at)orm a( a Ser+i&e 6PaaS8- and So)t,are a( a Ser+i&e 6SaaS8 re('e&ti+e%y. A re&ent Ber/e%ey re'ort (tated <$%oud $om'uting- the %onghe%d dream o) &om'uting a( a uti%ity- ha( the 'otentia% to tran()orm a %arge 'art o) the I* indu(try- ma/ing (o)t,are e+en more attra&ti+e a( a (er+i&e=. $%oud( aim to dri+e the de(ign o) the net generation data &enter( by ar&hite&ting them a( net,or/( o) +irtua% (er+i&e( 6hard,are- databa(e- u(erinter)a&e- a''%i&ation %ogi&8 (o that u(er( &an a&&e(( and de'%oy a''%i&ation( )rom any,here in the ,or%d on demand at &om'etiti+e &o(t( de'ending on their @oS 6@ua%ity o) Ser+i&e8 re;uirement( .
2
BINAYA K!AR PA"#I
1302001377
Need of Cloud Computing
*he need o) &%oud &om'uting &an be e'%ained ,ith the he%' o) an eam'%e. *he )o%%o,ing gra'h (ho,( the number o) u(er( ,ho %og on to the Au(tra%ian 'en ,eb 'age.
fig 1monthly page views of australian open official website
*he ('i/e( &orre('ond to the month o) anuary during ,hi&h the tournament i( going on. *he (ite remain( a%mo(t dormant during the re(t o) the year. It ,ou%d be ,a(te)u% to ha+e (er+er( ,hi&h &an &ater to the maimum need-a( they ,ont be needed during the re(t o) the year. *he &on&e't o) &%oud &om'uting &ome( to the re(&ue at thi( time. "uring the 'ea/ 'eriod- &%oud 'ro+ider( (u&h a( ?oog%e-Yahoo-!i&ro(o)t et&.&an be a''roa&hed to 'ro+ide the ne&e((ary (er+er &a'a&ity.
In thi( &a(e- In)ra(tu&ture i( 'ro+ided a( a (er+i&e6IaaS8 through &%oud &om'uting. 4i/e,i(e-&%oud 'ro+ider( &an be a''roa&hed )or obtaing (o)t,are or '%at)orm a( a (er+i&e. "e+e%o'er( ,ith inno+ati+e idea( )or ne, Internet (er+i&e( no %onger re;uire %arge &a'ita% out%ay( in hard,are to de'%oy their (er+i&e or human e'en(e to o'erate it . $%oud &om'uting o))er( (igni)i&ant bene)it( to I* &om'anie( by )reeing them )rom the %o,%e+e% ta(/ o) (etting u' ba(i& hard,are and (o)t,are in)ra(tru&ture( and thu( enab%ing )o&u( on inno+ation and &reating bu(ine(( +a%ue )or their (er+i&e(. 3
BINAYA K!AR PA"#I
1302001377
Green Computing ?reen &om'uting i( de)ined a( the (tudy and 'ra&ti&e o) de(igning - manu)a&turing- u(ingand di('o(ing o) &om'uter(- (er+er(- and a((o&iated (ub(y(tem((u&h a( monitor(- 'rinter((torage de+i&e(- and net,or/ing and &ommuni&ation( (y(tem(e))i&ient%y and e))e&ti+e%y ,ith minima% or no im'a&t on the en+ironment.C *he goa%( o) green &om'uting are (imi%ar to green &hemi(tryD redu&e the u(e o) haEardou( materia%(- maimiEe energy e))i&ien&y during the 'rodu&tF( %i)etime- and 'romote the re&y&%abi%ity or biodegradabi%ity o) de)un&t 'rodu&t( and )a&tory ,a(te. Re(ear&h &ontinue( into /ey area( (u&h a( ma/ing the u(e o) &om'uter( a( energye))i&ient a( 'o((ib%e- and de(igning a%gorithm( and (y(tem( )or e))i&ien&yre%ated &om'uter te&hno%ogie(.
*here are (e+era% a''roa&he( to green &om'uting-name%y •
Produ&t %ongeti+ity
•
A%gorithmi& e))i&ein&y
•
Re(our&e a%%o&ation
•
>irtua%i(ation
•
Po,er management et&.
Need of Green Computing In Clouds !odern data &enter(- o'erating under the $%oud &om'uting mode% are ho(ting a +ariety o) a''%i&ation( ranging )rom tho(e that run )or a )e, (e&ond( 6e.g. (er+ing re;ue(t( o) ,eb a''%i&ation( (u&h a( e&ommer&e and (o&ia% net,or/( 'orta%( ,ith tran(ient ,or/%oad(8 to tho(e that run )or %onger 'eriod( o) time 6e.g. (imu%ation( or %arge data (et 'ro&e((ing8 on (hared hard,are '%at)orm(. *he need to manage mu%ti'%e a''%i&ation( in a data &enter &reate( the &ha%%enge o) ondemand re(our&e 'ro+i(ioning and a%%o&ation in re('on(e to time+arying ,or/%oad(. Norma%%y- data &enter re(our&e( are (tati&a%%y a%%o&ated to a''%i&ation(- ba(ed on 'ea/ %oad &hara&teri(ti&(- in order to maintain i(o%ation and 'ro+ide 'er)orman&e guarantee(. nti% re&ent%yhigh 'er)orman&e ha( been the (o%e &on&ern in data &enter de'%oyment( and thi( demand ha( been G
BINAYA K!AR PA"#I
1302001377
)u%)i%%ed ,ithout 'aying mu&h attention to energy &on(um'tion. *he a+erage data &enter &on(ume( a( mu&h energy a( 2H-000 hou(eho%d( 20J. A( energy &o(t( are in&rea(ing ,hi%e a+ai%abi%ity d,ind%e(- there i( a need to (hi)t )o&u( )rom o'timi(ing data &enter re(our&e management )or 'ure 'er)orman&e to o'timi(ing )or energy e))i&ien&y ,hi%e maintaining high (er+i&e %e+e% 'er)orman&e. A&&ording to &ertain re'ort(-the tota% e(timated energy bi%% )or data &enter( in 2010 i( 11.H bi%%ion and energy &o(t( in a ty'i&a% data &enter doub%e e+ery )i+e year(.
"ata &enter( are not on%y e'en(i+e to maintain- but a%(o un)riend%y to the en+ironment. "ata &enter( no, dri+e more in &arbon emi((ion( than both Argentina and the Nether%and( . #igh energy &o(t( and huge &arbon )oot'rint( are in&urred due to ma((i+e amount( o) e%e&tri&ity needed to 'o,er and &oo% numerou( (er+er( ho(ted in the(e data &enter(. $%oud (er+i&e 'ro+ider( need to ado't mea(ure( to en(ure that their 'ro)it margin i( not dramati&a%%y redu&ed due to high energy &o(t(. Lor in(tan&e- ?oog%e- !i&ro(o)t- and Yahoo are bui%ding %arge data &enter( in barren de(ert %and (urrounding the $o%umbia Ri+er- SA to e'%oit &hea' and re%iab%e hydroe%e&tri& 'o,er . *here i( a%(o in&rea(ing 're((ure )rom ?o+ernment( ,or%d,ide to redu&e &arbon )oot'rint(- ,hi&h ha+e a (igni)i&ant im'a&t on &%imate &hange. Lor eam'%e- the a'ane(e go+ernment ha( e(tab%i(hed the a'an "ata $enter $oun&i% to addre(( the (oaring energy &on(um'tion o) data &enter( . 4eading &om'uting (er+i&e 'ro+ider( ha+e a%(o re&ent%y )ormed a g%oba% &on(ortium /no,n a( *he ?reen ?rid to 'romote energy e))i&ien&y )or data &enter( and minimi(e their en+ironmenta% im'a&t. 4o,ering the energy u(age o) data &enter( i( a &ha%%enging and &om'%e i((ue be&au(e &om'uting a''%i&ation( and data are gro,ing (o ;ui&/%y that in&rea(ing%y %arger (er+er( and di(/( are needed to 'ro&e(( them )a(t enough ,ithin the re;uired time 'eriod. Green Cloud computing i( en+i(ioned to a&hie+e not on%y e))i&ient 'ro&e((ing and uti%i(ation o) &om'uting in)ra(tru&ture- but a%(o minimi(e energy &on(um'tion. *hi( i( e((entia% )or en(uring that the )uture gro,th o) $%oud &om'uting i( (u(tainab%e. ther,i(e- $%oud &om'uting ,ith in&rea(ing%y 'er+a(i+e )rontend &%ient de+i&e( intera&ting ,ith ba&/end data &enter( ,i%% &au(e an enormou( e(&a%ation o) energy u(age. *o addre(( thi( 'rob%em- data &enter re(our&e( need to be managed in an energye))i&ient manner to dri+e ?reen $%oud &om'uting. In 'arti&u%ar- $%oud re(our&e( need to be a%%o&ated not on%y to (ati()y @oS re;uirement( ('e&i)ied by u(er( +ia Ser+i&e 4e+e% Agreement( 6S4A8- but a%(o to redu&e energy u(age.
H
BINAYA K!AR PA"#I
1302001377
Ar&hite&ture o) a green &%oud &om'uting '%at)orm
!ig " architecture of a green cloud computing environment
Ligure 2 (ho,( the high%e+e% ar&hite&ture )or (u''orting energye))i&ient (er+i&e a%%o&ation in ?reen $%oud &om'uting in)ra(tru&ture. *here are ba(i&a%%y )our main entitie( in+o%+ed: a) Consumers/Brokers: $%oud &on(umer( or their bro/er( (ubmit (er+i&e re;ue(t( )rom any,here
in the ,or%d to the $%oud. It i( im'ortant to noti&e that there &an be a di))eren&e bet,een $%oud &on(umer( and u(er( o) de'%oyed (er+i&e(. Lor in(tan&e- a &on(umer &an be a &om'any de'%oying a eb a''%i&ation- ,hi&h 're(ent( +arying ,or/%oad a&&ording to the number o) u(er( a&&e(ing it.
b) Green Resource Allocator : A&t( a( the inter)a&e bet,een the $%oud in)ra(tru&ture and
&on(umer(. It re;uire( the intera&tion o) the )o%%o,ing &om'onent( to (u''ort energye))i&ient re(our&e management:
BINAYA K!AR PA"#I
1302001377
#reen $egotiator : Negotiate( ,ith the &on(umer(Mbro/er( to )ina%iEe the S4A ,ith
('e&i)ied 'ri&e( and 'ena%tie( 6)or +io%ation( o) S4A8 bet,een the $%oud 'ro+ider and &on(umer de'ending on the &on(umer9( @oS re;uirement( and energy (a+ing (&heme(. In &a(e o) eb a''%i&ation(- )or in(tan&e- @oS metri& &an be H o) re;ue(t( being (er+ed in %e(( than 3 (e&ond(.
%ervice Analyser : Inter'ret( and ana%y(e( the (er+i&e re;uirement( o) a (ubmitted re;ue(t
be)ore de&iding ,hether to a&&e't or re5e&t it. #en&e- it need( the %ate(t %oad and energy in)ormation )rom >! !anager and nergy !onitor re('e&ti+e%y.
Consumer &rofiler : ?ather( ('e&i)i& &hara&teri(ti&( o) &on(umer( (o that im'ortant
&on(umer( &an be granted ('e&ia% 'ri+i%ege( and 'rioriti(ed o+er other &on(umer(
&ricing : "e&ide( ho, (er+i&e re;ue(t( are &harged to manage the (u''%y and demand o)
&om'uting re(our&e( and )a&i%itate in 'rioriti(ing (er+i&e a%%o&ation( e))e&ti+e%y.
'nergy (onitor : b(er+e( and determine( ,hi&h 'hy(i&a% ma&hine( to 'o,er onMo)). %ervice %cheduler : A((ign( re;ue(t( to >!( and determine( re(our&e entit%ement( )or
a%%o&ated >!(. It a%(o de&ide( ,hen >!( are to be added or remo+ed to meet demand.
)( (anager : Kee'( tra&/ o) the a+ai%abi%ity o) >!( and their re(our&e entit%ement(. It i(
a%(o in &harge o) migrating >!( a&ro(( 'hy(i&a% ma&hine( Accounting : !aintain( the a&tua% u(age o) re(our&e( by re;ue(t( to &om'ute u(age &o(t(.
#i(tori&a% u(age in)ormation &an a%(o be u(ed to im'ro+e (er+i&e a%%o&ation de&i(ion(.
&8 VMs: !u%ti'%e >!( &an be dynami&a%%y (tarted and (to''ed on a (ing%e 'hy(i&a% ma&hine to meet a&&e'ted re;ue(t(- hen&e 'ro+iding maimum )%eibi%ity to &on)igure +ariou( 'artition( o) re(our&e( on the (ame 'hy(i&a% ma&hine to di))erent ('e&i)i& re;uirement( o) (er+i&e re;ue(t(. !u%ti'%e >!( &an a%(o &on&urrent%y run a''%i&ation( ba(ed on di))erent o'erating (y(tem en+ironment( on a (ing%e 'hy(i&a% ma&hine. In addition- by dynami&a%%y migrating >!( a&ro(( 'hy(i&a% ma&hine(- ,or/%oad( &an be &on(o%idated and unu(ed re(our&e( &an be 'ut on a %o,'o,er (tate- turned o)) or &on)igured to o'erate at %o,'er)orman&e %e+e%( 6e.g.- u(ing ">LS8 in order to (a+e energy.
d8 Physical Machines: *he under%ying 'hy(i&a% &om'uting (er+er( 'ro+ide hard,are in)ra(tru&ture )or &reating +irtua%i(ed re(our&e( to meet (er+i&e demand(.
7
BINAYA K!AR PA"#I
1302001377 !a/ing &%oud &om'uting more green
!ain%y three a''roa&he( ha+e been tried out to ma/e &%oud &om'uting en+ironment( more en+ironmenta% )riend%y. *he(e a''roa&he( ha+e been tried out in the data &entre( under e'erimenta% &ondition(. *he 'ra&ti&a% a''%i&ation o) the(e method( are (ti%% under (tudy. *he method( are:
•
Dynamic Voltage frequency scaling technique(DV!) : +ery e%e&troni& &ir&utory ,i%%
ha+e an o'erating &%o&/ a((o&iated ,ith it. *he o'eratin )re;uen&y o) thi( &%o&/ i( ad5u(ted (o that the (u''%y +o%tage i( regu%ated. *hu(- thi( method hea+i%y de'end( on the hard,are and i( not &ontro%%aba%e a&&ording to the +arying need(. *he 'o,er (a+ing( are a%(o %o, &om'ared to other a''roa&he(. *he 'o,er (a+ing( to &o(t in&urred ratio i( a%(o %o,. •
Resource allocation or "irtual machine migration techniques : In a &%oud &om'uting
en+ironment-e+ery 'hy(i&a% ma&hine ho(t( a number o) +irtua% ma&hine( u'on ,hi&h the a''%i&ation( are run. *he(e +irtua% ma&hine( &an be tran()ered a&ro(( the ho(t( a&&ording to the +arying need( and a+aia%b%e re(our&e(.*he >! migration method )o&u((e( on tran()erring >!( in (u&h a ,ay that the 'o,er in&rea(e i( %ea(t. *he mo(t 'o,er e))i&ient node( are (e%e&ted and the >!( are tran()ered a&ro(( to them. *hi( method i( dea%t in detai% %ater. •
Algorithmic a##roaches: It ha( been e'eriment%y determined that an idea% (er+er
&on(ume( about 70 o) the 'o,er uti%i(ed by a )u%%y uti%i(ed (er+er. 6See )igure 38.
!ig * &ower consumption under different work loads.
O
BINAYA K!AR PA"#I
1302001377
(ing a neura% net,or/ 'redi&tor-the green (&hedu%ing a%gorithm( )ir(t e(timate( re;uired dynami& ,or/%oad on the (er+er(. *hen unne&e((ary (er+er( are turned o)) in order to minimiEe the number o) running (er+er(- thu( minimiEing the energy u(e at the 'oint( o) &on(um'tion to 'ro+ide bene)it( to a%% other %e+e%(. A%(o-(e+era% (er+er( are added to he%' a((ure (er+i&e%e+e% agreement. *he bottom %ine i( to 'rote&t the en+ironment and to redu&e the tota% &o(t o) o,ner(hi' ,hi%e en(uring ;ua%ity o) (er+i&e.
BINAYA K!AR PA"#I
1302001377 >! !igration
*he 'rob%em o) >! a%%o&ation &an be di+ided in t,o: the )ir(t 'art i( admi((ion o) ne, re;ue(t( )or >! 'ro+i(ioning and '%a&ing the >!( on ho(t(- ,herea( the (e&ond 'art i( o'timiEation o) &urrent a%%o&ation o) >!(. 'timiEation o) &urrent a%%o&ation o) >!( i( &arried out in t,o (te'(: at the )ir(t (te' ,e (e%e&t >!( that need to be migrated- at the (e&ond (te' &ho(en >!( are '%a&ed on ho(t( u(ing !BL" a%gorithm. e 'ro'o(e )our heuri(ti&( )or &hoo(ing >!( to migrate. *he )ir(t heuri(ti&Single Threshold (ST)- i( ba(ed on the idea o) (etting u''er uti%iEation thre(ho%d )or ho(t( and
'%a&ing >!( ,hi%e /ee'ing the tota% uti%iEation o) $P be%o, thi( thre(ho%d. *he aim i( to 're(er+e )ree re(our&e( to 're+ent S4A +io%ation due to &on(o%idation in &a(e( ,hen uti%iEation by >!( in&rea(e(. At ea&h time )rame a%% >!( are rea%%o&ated u(ing !BL" a%gorithm ,ith additiona% &ondition o) /ee'ing the u''er uti%iEation thre(ho%d not +io%ated. *he ne, '%a&ement i( a&hie+ed by %i+e migration o) >!( . *he other three heuri(ti&( are ba(ed on the idea o) (etting u''er and %o,er uti%iEation thre(ho%d( )or ho(t( and /ee'ing tota% uti%iEation o) $P by a%% >!( bet,een the(e thre(ho%d(. I) the uti%iEation o) $P )or a ho(t goe( be%o, the %o,er thre(ho%d- a%% >!( ha+e to be migrated )rom thi( ho(t and the ho(t ha( to be (,it&hed o)) in order to e%iminate the id%e 'o,er &on(um'tion. I) the uti%iEation goe( o+er the u''er thre(ho%d- (ome >!( ha+e to be migrated )rom the ho(t to redu&e uti%iEation in order to 're+ent 'otentia% S4A +io%ation. e 'ro'o(e three 'o%i&ie( )or &hoo(ing >!( that ha+e to be migrated )rom the ho(t.
!inimiEation o) !igration( 6!!8 migrating the %ea(t number o) >!( to minimi(e
migration o+erhead. #ighe(t Potentia% ?ro,th 6#P?8 migrating >!( that ha+e the %o,e(t u(age o)
$P re%ati+e%y to the re;ue(ted in order to minimi(e tota% 'otentia% in&rea(e o) the uti%iEation and S4A +io%ation Random $hoi&e 6R$8 &hoo(ing the ne&e((ary number o) >!( by 'i&/ing them
a&&ording to a uni)orm%y di(tributed random +ariab%e.
10
BINAYA K!AR PA"#I
1302001377
$%#erimental !etu#
A( the targeted (y(tem i( a generi& $%oud &om'uting en+ironment- it i( e((entia% to e+a%uate it on a %arge(&a%e +irtua%i(ed data &enter in)ra(tru&ture. #o,e+er- it i( di))i&u%t to &ondu&t %arge(&a%e e'eriment( on a rea% in)ra(tru&ture- e('e&ia%%y ,hen it i( ne&e((ary to re'eat the e'eriment ,ith the (ame &ondition( 6e.g. ,hen &om'aring di))erent a%gorithm(8. *here)ore- (imu%ation( ha+e been &ho(en a( a ,ay to e+a%uate the 'ro'o(ed heuri(ti&(. *he $%oudSim too%/it ha( been &ho(en a( a (imu%ation '%at)orm a( it i( a modern (imu%ation )rame,or/ aimed at $%oud &om'uting en+ironment(. In &ontra(t to a%ternati+e (imu%ation too%/it( 6e.g. Sim?rid- ?andSim8- it (u''ort( mode%ing o) ondemand +irtua%iEation enab%ed re(our&e and a''%i&ation management. It ha( been etended in order to enab%e 'o,era,are (imu%ation( a( the &ore )rame,or/ doe( not 'ro+ide thi( &a'abi%ity. A'art )rom the 'o,er &on(um'tion mode%ing and a&&ounting- the abi%ity to (imu%ate (er+i&e a''%i&ation( ,ith +ariab%e o+er time ,or/%oad ha( been in&or'orated. *here are a )e, a((um'tion( that ha+e been made to (im'%i)y the mode% o) the (y(tem and enab%e (imu%ationdri+en e+a%uation. *he )ir(t a((um'tion i( that the o+erhead o) >! migration i( &on(idered a( neg%igib%e. !ode%ing the &o(t o) migration o) >!( i( another re(ear&h 'rob%em and i( being &urrent%y in+e(tigated . #o,e+er- it ha( been (ho,n that a''%i&ation o) %i+e migration o) >!( &an 'ro+ide rea(onab%e 'er)orman&e o+erhead. !oreo+er- ,ith ad+an&ement( o) +irtua%iEation te&hno%ogie(- the e))i&ien&y o) >! migration i( going to be im'ro+ed. Another a((um'tion i( that due to un/no,n ty'e( o) a''%i&ation( running on >!(- it i( not 'o((ib%e to bui%d the ea&t mode% o) (u&h a mied ,or/%oad . *here)ore- rather than (imu%ating 'arti&u%ar a''%i&ation(- the uti%iEation o) $P by a >! i( generated a( a uni)orm%y di(tributed random +ariab%e. In the (imu%ation( ,e ha+e de)ined that S4A +io%ation o&&ur( ,hen a >! &annot get amount o) !IPS that are re;ue(ted. *hi( &an ha''en in &a(e( ,hen >!( (haring the (ame ho(t re;uire higher $P 'er)orman&e that &annot be 'ro+ided due to &on(o%idation. *o &om'are e))i&ien&y o) the a%gorithm( ,e u(e a &hara&teri(ti& &a%%ed S4A +io%ation 'er&entage- or (im'%y S4A +io%ation- ,hi&h i( de)ined a( a 'er&entage o) S4A +io%ation e+ent( re%ati+e%y to the tota% number o) mea(urement(.
A data &enter that &om'ri(e( 100 heterogeneou( 'hy(i&a% node( ,a( (imu%ated. a&h node i( mode%ed to ha+e one $P &ore ,ith 'er)orman&e e;ui+a%ent to 1000- 2000 or 3000 !i%%ion In(tru&tion( Per Se&ond 6!IPS8- O ?b o) RA! and 1 *B o) (torage. A&&ording to thi( mode%- a ho(t &on(ume( )rom 17H ,ith 0 $P uti%iEation and u' to 2H0 ,ith 100 $P uti%iEation. a&h >! re;uire( one $P &ore ,ith 2H0- H00- 7H0 or 1000 !IPS- 12O !B o) RA! and 1 ?B o) (torage. *he u(er( (ubmit re;ue(t( )or 'ro+i(ioning o) 20 heterogeneou( >!( that )i%%( the )u%% 11
BINAYA K!AR PA"#I
1302001377
&a'a&ity o) the (imu%ated data &enter. a&h >! run( a ,eba''%i&ation or any /ind o) a''%i&ation ,ith +ariab%e ,or/%oad- ,hi&h i( mode%ed to &reate the uti%iEation o) $P a&&ording to a uni)orm%y di(tributed random +ariab%e. *he a''%i&ation run( )or 1H0-000 !IPS that e;ua%( to 10 minute( o) ee&ution on 2H0 !IPS $P ,ith 100 uti%iEation. Initia%%y- >!( are a%%o&ated a&&ording to the re;ue(ted &hara&teri(ti&( a((uming 100 uti%iEation. a&h e'eriment ha( been run 10 time( and the 're(ented re(u%t( are bui%t u'on the mean +a%ue(.
!imulation Results
Lor the ben&hmar/ e'erimenta% re(u%t( ,e ha+e u(ed a Non Po,er A,are 6NPA8 'o%i&y. *hi( 'o%i&y doe( not a''%y any 'o,er a,are o'timiEation( and im'%ie( that a%% ho(t( run at 100 $P uti%iEation and &on(ume maimum 'o,er. *he (e&ond 'o%i&y a''%ie( ">LS- but doe( not 'er)orm any ada'tation o) a%%o&ation o) >!( in runtime. Lor the (imu%ation (etu' de(&ribed abo+e- u(ing the NPA 'o%i&y %ead( to the tota% energy &on(um'tion o) .1H Kh- ,herea( ">LS a%%o,( de&rea(ing thi( +a%ue to G.G Kh. *he (imu%ation re(ut( o) +ariou( 'o%i&ie( are e'%ained in the net (e&tion(.
12
BINAYA K!AR PA"#I
1302001377
nergy $on(um'tion and S4A +ioa%tion o) S* 'o%i&y
ig &'simulation results of ! #olicy
*o e+a%uate S* 'o%i&y ,e &ondu&ted (e+era% e'eriment( ,ith di))erent +a%ue( o) the uti%iEation thre(ho%d. *he (imu%ation re(u%t( are 're(ented in Ligure G. *he re(u%t( (ho, that energy &on(um'tion &an be (igni)i&ant%y redu&ed re%ati+e%y to NPA and ">LS 'o%i&ie( by 77 and H3 re('e&ti+e%y ,ith H.G o) S4A +io%ation(. *hey (ho, that ,ith the gro,th o) the uti%iEation thre(ho%d energy &on(um'tion de&rea(e(- ,herea( 'er&entage o) S4A +io%ation( in&rea(e(. *hi( i( due to the )a&t that higher uti%iEation thre(ho%d a%%o,( more aggre((i+e &on(o%idation o) >!(ho,e+er- by the &o(t o) the in&rea(ed ri(/ o) S4A +io%ation(.
nergy &on(um'tion and S4A +io%ation( o) other 'o%i&ie(
e ha+e &om'ared !! 'o%i&y ,ith #P? and R$ 'o%i&ie( +arying ea&t +a%ue( o) the thre(ho%d( but 're(er+ing G0 inter+a% bet,een them. *he re(u%t( 6Ligure( H Q 8 (ho, that the(e 'o%i&ie( a%%o, the a&hie+ement o) a''roimate%y the (ame +a%ue( o) energy &on(um'tion and S4A +io%ation(. herea( the number o) >! migration( 'rodu&ed by !! 'o%i&y i( redu&ed in 13
BINAYA K!AR PA"#I
1302001377
&om'ari(on to #P? 'o%i&y by maimum o) H7 and G0 on a+erage and in &om'ari(on to R$ 'o%i&y by maimum o) G and 27 on a+erage.
ig 'energy consum#tion of *ifferent #olicies
ig+'!,A "iolations of *ifferent #olicies un*er *ifferent threshol*s
1G
BINAYA K!AR PA"#I
1302001377
Com#arison -ith res#ect to CP. utilisation
ig 'com#arison of *iff #olicies un*er *ifferent -orkloa*s -ith res#ect to c#u utilsation
A( &an be (een-the Non Po,er A,are 6NPA8 'o%i&y ,hi&h i( &urrent%y being )o%%o,ed re(u%t( in higher 'o,er %o((-e+en though it doe(nt re(u%t in S4A +io%ation(. *he o'tima% &a(e o&&ur( ,hen !! 'o%i&y under a ,or/%oad o) 3070i( u(ed.
$on&%u(ion
A''%ying green te&hno%ogie( i( high%y e((entia% )or the (u(tainab%e de+e%o'ment o) &%oud &om'uting. ) the +ariou( green methodo%ogie( en;uired- the ">LS te&hno%ogy i( a high%y hard,are oriented a''roa&h and henn&e %e(( )%eib%e. *he reu(%t o) +ariou( >! migration (imu%ation( (ho, that !! 'o%i&y %ead( to the be(t energy (a+ing(: by O3- and 23 %e(( energy &on(um'tion re%ati+e%y to NPA- ">LS and S* 'o%i&ie( re('e&ti+e%y ,ith thre(ho%d( 3070 and en(uring 'er&entage o) S4A +io%ation( o) 1.1D and by O7- 7G and G3 ,ith thre(ho%d( H0 0 and .7 o) S4A +io%ation(. !! 'o%i&y %ead( to more than 10 time( %e(( >! migration( than S* 'o%i&y. *he re(u%t( (ho, )%eibi%ity o) the a%gorithm- a( the thre(ho%d( &an be ad5u(ted a&&ording to S4A re;uirement(. Stri&t S4A 61.118 a%%o, the a&hie+ement o) the energy &on(um'tion o) 1.GO Kh. #o,e+er- i) S4A are re%aed 6.8- the energy &on(um'tion i( )urther redu&ed to 1.1G Kh. Sing%e thre(ho%d 'o%i&ie( &an (a+e 'o,er u'to 20-but they a%(o &au(e a %arge number o) 1H
BINAYA K!AR PA"#I
1302001377
S4A +io%ation(. ?reen (&hedu%ing a%gorithm( ba(ed on neura% 'redi&tor( &an %ead to a 70 'o,er (a+ing(. *he(e 'o%i&ie( a%(o enab%e u( to &ut do,n data &entre energy &o(t(- thu( %eading to a (trong-&om'etiti+e &%oud &om'uting indu(try. nd u(er( ,i%% a%(o bene)it )rom the de&rea(ed energy bi%%(.
1
BINAYA K!AR PA"#I
1302001377 Re)eren&e(
1. <nergy e))i&ient management o) data &entre re(our&e( )or &%oud &om'uting:A
+i(ion-ar&hite&tura% e%ement( and 'en $ha%%enge(= Ra5/umar Buyya-Anton Be%og%aEo+-ema% Aba,a5y &roc. of +th ''' nternational %ymposium on Cluster Computing and the #rid -CC#rid "+/ - Rio "e aneiro- BraEi%- !ay 200. ".
&om'uting= *roung >inh *roung "uy-Yu/inori Sato-Ya(hu(hi Inogu&hi ''' 0plore, (arch "1 *. ,,,.,i/i'edia.&omMgreen&om'uting. . ,,,.ibm.&omMde+e%o'er,or/(M,eb('hereMEone(Mhi'od(
17