co U-2 com i ~ cloud ampusting and sesrvice models ¢ | cloud computing defined as the Practice of ice OF US! ing a netuosk oF verole sewers hos kd on tntrnet + lo Stowe, manage & precess dato oud computing types : ge | a ; aby Sof ware as a svivie (Saas) & Plat , Plat formas a Service (.paas) 3) infrasivuctare a i 3 Ink a Sewice (1aas) 4 Any wrirg [Everything as aso ee co Function as 0 setvice (Faas) » Saas? i a . ob $Aa' Saas is a way OF delivering eewvices. and applica tons over «teenth ingead oF lastalling & maintaining sof wore we cimply acess te viow Internet: gaas provides comple k Softaare soln that we purchase on payasyag? pasis fom doud sewia provider Saas 01s called as web pared software on-demand software ov hosted Sof two gaye costet fective, Reduced tame, nccesstbility Automatic updates, coatabeility bis ¢ Limit customization, security toncan , intaner dependent * A Paas% Paas provides a platform environment to qulow dsvelapers to pulid apphaaton® gy services ovey the ingore b~ eb vous]: Hosta gn cloud @ accerted by wsers via hel W Adve cost ef ecive- ciple & coavinient fov wsers pis: cimited contro| over flex! iliby, dependent on provider: or IRAs? (a10% ig a suwvice model ynat delivers Computer lnfaswructuve On asis 10 support various operations: on qursowad bi aos provided to devices, doctavares and websewvers” aos also yenown as hardware as a servi” adv. cost ef lecHve, onsite hosting» cetusity ,maintanend - ois, gecuTity conus, Limited ACCES, Limited Control overt Inhowtucture a Xaasi xaos also Known OS everything af a swice- Now o days most oF cloud provides 40 provide anything ac a sovvie- Adve scalability (rerivitity , wse el feckive Bee Uaiied clexibility, timited salequation, Oenendence on Nhe Prov erfooss Faas ; ) OS provide. < o pete fe wen OF uustomets +2 develoP, pute ceun and aerioy Whe wae ov entire aoniiation 4 ac function” tr G(GWS users to develor she wde ¥ upload it ang aout mainknand- ie if ‘implement 19 gervestless contain” in foas “i pay for number oF exeuation HME append pave Highly suaiaWle, cosy - cde, code simpli Hate? aise coi Stare Latenc4, seutrity (nuns, umitd oo Cloual Platt forms puplic houd ira vange oF womputing sewices jnat are microsoft nzuve, Google tol tf | atari ty * o creved by piarforms suth as amazon web SEAN ces caw), cloud platiowm cyte) and making them availapie ove7! fakoinet aminihraA by wiv puplic coud opencckd by pay ~O5- Pe aie modal 4 hh ea croud service povidel” in public cloud same storage! used by multiple Woe ak we same time public cloud owned, managed and opena ted by a government ong! 22 Hons pduy Lowcost? | vocation dependent § Shosavle = ‘( storage a) a equetimes , Quicky Lass setups, Bussiness Agilibye seutaiot lity & Retiata ty = a Dist Low sewmityé » petormance , Legs customizable + SeAvice ovienkd pron eceure + Senvice orien ea prchivectuve wa stage in evolution of otpplication everopment + integration - te defines away 40 make software components yewane wing 1 1 uses coramon eure FOS Pe icaton etonaavds to speedup Y streamline dine cee chawactoristics ¢ provide (oosely couple SenMies’ provides towatlon tvanspavency with S andoroperablliby blw the Services: provides verter swauawilib ond availability * potovides methods {or ceil neusabillly , sevice composition senvice aisw@ovey? Sei consumption | Sewice Inkgrat or).Pere neste eeteee geese Nee eee NEe Eee geet EEE REE EEeEECEMEEEEceEuEEECCCNEEEE Co uEeEC EEE ue major voles of SON? su seawvice pAOvIdeT § envice provide) Is mm ainainet of sonvice @ tHE ovgarizin shot makes available one ov move services a _y seviee Consumert The soma consumer (ar oat tne soe mero data ta ane vegiony ard develeP ine veyuived cliente component? tovind «ite sevice er {cewleoe poder | : ems of SOME ewig vi “s A Os csandandned sevice ze) Eye Fewer) tory envi sain irra i {ew wa | cay voote coupling oe (ao nipstvacten + entation] a) forte] eo Reusaeiltag + — (s) futonomy * giswvenan' Wiby & composai (iby . wb pays sowie geusability, availawility, etiability, Gacy rma ronands platfominaerendert Oise Hight investment pighovernead, ConnPle: serie. nanagerert - Applic? vad by Avmys | Niv-forees , healthcare leliverties » MUdUms * programming CU on fws e Azures Aus? amazon wer cenvices HIS provided 4 amazon that uses digit ry infvarcactuse to provide different IT vesouras available on aemawd 1% provides services suchas Laas, fags, SOS © ion wes tel serterpri se 4o expand A small manuta ctuving ovganizali utes: aneiv yusiness* 0 Large erierernse spread across dhe glove @ utilize pws 10 deliver pou ag-Hou- Gee fave EF iexwility, cost et ectlve, scalability, security: eager ) Qo ‘por! Gornpatieg) (pH as you ae 7} a [ nekwoen yf ~(ecgrane — oe iawn) ese ” (nw)ca instants clastic Compute cloud vplead® fF pov ct (sr0a¢] Sa Bucket S , ees cimnple SoG sowie vowel | a| Aus Seuuviy |__—___— ave pus Pitre paicrosoft ADULNE = pueels & micvosoF t's ‘oud platform jut like Jeg” mite plariwn though) which we uk mivosoft yerourdds- que if alio simple & (ogt -e FREE anaiyt wal Y prousing of ota, we wiang of O ization « will provide us viv ety completely ban tp in bussiness cost effectives pole e , Cody eackuy tual machine -fost te don process oF virtual monitoring 4jools: How pave (an vo copia bess # $ operational , BOY 4o implement z _ ware fomany vate g Recovey= | Berta KUNE ves micosoft pure fowl ase icnosofe eb compute J) comtainels Wy internet of things cay Nerworkny 1% dabobass toy seenriby + identity: @ Alt cognitive sowids ce srov0ge 7) wetmatav? La vara + Analy tes [exeogvarny] — Region oes —L een [avails zones ~ gpaita center | Quien 7 AFC- [ec] “piuve Fabvic Controle)we Eyalano lution OF s forage tKchnologi ies ured ato break Advancement 1 0 data storage sechnolagy have Oct has cK S| _neck Speed and tre gunaunt oF infomation that is saved ovaily been stadily gyowing: SHov0ge yechnology ‘evolved fvom non- ink networked StovOge- pegundart Anogot indevendent visks (wave)? Tis arwolgy was devebred wo odavess bre costs perlormane canal qvailabitity veasivererts ° wa in all gtovagt architecture such OS pas, SAN and $0 on ect atiached storage coAs): juigent inkmal S08 fo incite £ date pwectly attacked to a cotver ov O group owe! of ewes tn a clus erovage AVEO petwerk C5AND Te V6 adedicated high porformanc., Flere channel vers & etovage- biock- level communication werueen Ser availaility, perforant & Lost benehits- a cedicard stovoge for file serving cipplicadns jalest evolutions in $ tocal oy wide aed. network: network faciitott SAN offers stale! Sreod (Nas): 1b 1S network Auioched ov age Ov cbitect. protoccl san Cip-SAND= one of “the | $0. inxenet ommunica Hor across 6 block revel U evolution of storage EE cevies: a au -usporive, 2006- cloudconputig (464 - FlopP4 disk - wx storage Aedes In “in cloud va computing cloud t0rag' cis O virtual locker whee we can poued sewer Ike google asive, ets copied overt the inkanet into odata dreds of the seyuers enawn as data. vill not one sive, é ¢ cloud anat file 4 anies have huni across multifle (ocations: $0 our ¢ cloud stovage system 50 Ut has a queater availability joe features OF of ose mainenane * Automate sys o hash Loge ne work ACCESS *Ce TIE eX Stovage systems in the cloud: her cote “Inve types of Storage sysierns in em ¥ File Based stovagesystn? cloud: & BlOKK- Based Storage SHS ¥ object Based etorage SyKr ess: Hard drives ae block based storage systems 0% like windows “ina® acually sees a N07 dig dlsive- A drive on which wean eveate 0 volun and then we can partition nat volume & format tem ext 1 a system ras 100548 oF voumes We van parfition vt inl? 00 46% 200.48 card (NIC) a network interface fov tocol “¢ and ‘p' drives: a network insnis we actually on one Filepased $10 4ne network nec ting t{hvough Hf we ue going ovel Fee: jovage seer (NAS) NAS devas qesyskm won auess arached & ynis a He uploads objec using & web browse object stow contain f the APTS” ned t0 gowniaad some anen We ages: 7 4 prods an omgect ris Wes EX? when yo aenages jut OF ang ady: 9¢ pis: pate Long aor contain’! ven 5 toe protocols wit ne nttp conneck any “4 worse, we tring that websit contains aiabiliby pelexiolliby RENN, cenwns yeqpdve olectwiciby and pepe! utticient an physical devid~ gimplet dete migvations: intanet > ig not $ * gute systtt’ and a aatabase ¢ m ig wasicaly awsy of onary” qne file cyst ond disk ore qvouped sto directories’ ne vetnieval of files 9) ane files ina Stovage and hells int ey ave reqpived’ filenaming - giving ace ess ules ele ext NTPSC OC El File sy J Mast; File Directory Filesysteey iy Fi de [i] Ulertile divectory filesx potabase Management System 3 cofwore that manages +6 conection Of aa ef fecnvelg ne aote pems is basically system tt 16 wed 4ov stoving & yetviewng yelakd data- pootec ting {te also provides pooper Sequaity measures for yeiatioral aiger® yackyP yom unauthovind access” can be fetched by SQL queries ard In Dems date pecovery C 1A tt also provides mechanism {or dato ext orade, mgs And user) patowase — le of ie ——08"5 (rooted ~ o 3 Dems eh [pata tase) ae pistvibuttd Filesystem Cpr) ¢ A dictvibuted fitecyson isa fie syeem 4 on multiple x multiple (ova. bons” mmers to alless 41les e of OFS is 4o atlow using a commen flesyston at is distributed file sevvers 0 fvom any network: theses to share their data It allows progvas The main puypos and vesouras by Drs nas two components 3 twanspavency: Ib 16 achieved thvough the voume Spaw component: Location Sb 16 done 4rrough @ ile septication component - Reaundan 44 reotuves Of DES: 1) Tyansper al ch Tyanspastenly abswucture tvansparency & Ui, Naess tains pavenc ye iit) Naming ~wans; 2 ¢ ing Avanspavenuy? iv) Replication svansparencys OO) voeq Mobility s tr wi ty will automatically bvi ings +he usen's hor ne «perform z ona: A, High setiability: c s no , 4 Simpidty and ease of wee By CALA Inte : Integrity ¢ 6 High availa bilidys GM Secusit oe AU Secuvity & Scalability: Wye sto, Heterogeneity z Applications s cb WESC Chetwosk Filesystenn) () Hado loop ciese Cc oS COmmor ry > Internet Fle Susteen) Smeg 1B ( Servers Message eI ) Neloore lock) .posKing Of OFS: —— so oe Ww? wags 1° whith prs neementa jondalone DFS namespau: A standalone DFs can onlty be acquire on thee computers On sahigh it 16 Crean te doesn't provide fault liberation & can't pe tinked 40.04 other OFS romain based OFS namespet® i stows configuration active divectory” Ady: A\wows multiple vas poactess 45tONe dato: provides avanspavency, innproves access Hime, = peork eF dene a = ma ote avaiwavill ty of File bic: Comp wate, ceurity, 1066 of mesag Genew:| poralle! File systema oes woo Cludel Filesystem tt provides conwurent puess too single file syster fyom multiple nodes 1b is Filesyskm used to dighibute & s yoraqedevice s manage aote acwoss rnultiple senvers & iS implemented in mang righ pevbennde ups has forges ties close. seplicaton, polity posed stowage managrenE ard musi site operations: It descv ies all povauel file sysion pecause GPFS dato is soon N40 piocks ard oa acvoss Muldple disks nan anag~ neve ave two main types of GPRS cy cloud posed povaiel Filesystem uy Stora pavaliel filesyse™ Adve Date antegriv4, pata se wesity, pisastey yecovety” Filesyskim client (Compute) |rne st = ae __ Fabric _) aes server) af a 7 al - fe fi re y ise Low scalability , Less pertormane F ~ ee - File system) | cilesyskm — _client _IEE ESTE WOOO ES Google File System Google Inc a lever distvi bused ped tne goal file system CAFO A scalable itvibuted Gi File syskm (pFs) 4o meet company’s growing dato Proasging needs: ges oes foult olevance , dependablis , seatlabiltf ravaciabiltl and petfownanu 10 big networks | connect nodes” aes madeupof a num of Storage sys ¢ commodity hardware outs: QPS yedued of commercially availabe seivers ars manages 40 types oF dato. namely Grs 1b made 10 meat google’s huge cluset rrepaivements al Chunk SesVers * ars consiott of Single master and seve punk sewer by timed “the Masts communicates with each C eanit bear message” components of AFS* Ges clients < comput te cemves aS clus 4 application used +0 yauest files” Jen's. (oordinasor” 1, Wook horses” ‘Thy Keep 6yMe- form chunk Sowers: Fs mastyl sewer. GES ChUAK Servers ney Ore ne biFs sined File Chunke- master doesn't Ye ive any MS [soca yecstuweso€ 4FS* ~Gips mast Fault tolerant File chunk mare! Hignavai auillly rere | chunk engual dotavephaation | [eite! chunk! | pilet chunks ebiident aata ouely cilea : (unk namespa ie monogurent \ Filed anid WolEl¥G * a CFilegchunknamesoace) Adyze Hi pactessiblltty data Zhunk location mappin #4] pings Exugsive tnvoughnt A Tiitename,chunk inde) GES Client Capplication) + pependawle storing pile Hot the pest fit for em files Lnabie to type OF yandom All } CoWwect auess to DATA) Ges Chunk Sewers > Lola file syste aecloud Fesourc mona. wei Policies gm gement and scheduling? & Mechanisms ov verqurce management 2 Polley typical Y vekrs to -the principal guiding deaionsard wk mechani recast me a nt, in Td Geomine eben mecranisms ic @ guiding principe 1? - cb Admission Contvol gement poiies can be loosely grouped Into five Ga @ capacity allocation & load balancing 4) Energy optimization .% Quality of senice (Gos) ° Load balandng qenergy opkmraton are policies ancourered “WE | some ai€tcultles as the one Four basic mechanisms fov implementation of vesoure management policles aves ety) CoMtvol theory: (ad Machine Learning: <4 utility yased + ca Mav kee ovienkd /economic mechanism = ppptuations of | tontol Hey Jo yask scheduling n a Cloy cloud conivol tneory Nas ban used 10 design adaptive vesouva. managment Gr many Classes of applications ,induding powet management ,4ask scheduling Gos ee in web Senvers and load baalan ang qe feedback wontol varnant sysem model Youn components of a (ontval system: hy The inputs ¢ Gil, ThE Commvol system Components? Gilly The outputs? in these. methods assumes a linear ime ~ and a closed - 1900p comoley dy eed oak and Stability oudput __ distor vedic ane pial fe “eo ‘Lew st canta Tot wee) a eacegck) 5S ave ane weighting factors of (he Peformance jndlexasic sth i — cheduling +9 cloud computing: cud computing madeupof larg num of data pumeous physical machines chost) gach host vans several virtual machines Sateen yser tasks with vartying quality of cowie C405) ——) aisk-t : haus! J coud Bxoicen cloud - informa tea 7 5 | , Soul ocheduling Atoor'tHam ye Theta Tr 4OSES deve ave ge exceed using WW viv vm" anured 1000 fouow FLFS pain gis y ver OL ne evagnowd a suppose tual machines iciple- aw which ar in ders OF wand m= Ven Nema, V3» sthe num of tasks » tases Wave yurtouy len germ % CPU 4ime- y element ot +ask scheduting pros” aucnitectuve © uct contra les . cloud prokeris stavility of two (evel Resour alla te A two level contvol averitectuve appiication convo (ers & clo ey of controttey save prOMTAEL % Wye Applic e> ee n (weary) [vale a —— - enonitey monitor peciton oecifon Actuator Actuator orm = [21S choid — bared policter exnd iysequential clecition qwo polrrnare ud 6 itosear Scertarbje Computing oven Takin - Techngyl fy netwop-vasel Sustms” \ Sytem emedely fox dis iba CC WA Secu yt enonyerseney * WEY pap ‘ re Virtudlizotion Clevel s ,sschere, tol, mnechani sm) * Y f ® for vata conker tutomotion Virtual c(cstees « ia ATC Systems ) wv (Pap Network > datagras AW) Suiebao » ( servlets enolutiono F HTC g HPC syste Cornpuching) Paradigns: a Centvalized Computing + G—-Savu—c c prouwov —_ eee a hs cr rumors J a (==) leet, a £ Povalie! computing 5 vishiputed computa + cloud computing ; elent— Tries { pegrers of parallelions s Bit level poaotelisn 2 ,unstacton level faralleliomn?, oada (evel pavaletioms, elias , job leve | poucallelime task level pare vivo V2O0Un threading echrotogie!« multi cove, . pl SOOO erode upot muttile eves uittina cirgle Orererion 11! w such a4 pe (dual, quad) ete, Proeupy fo cawy out egnigln in poraiiel 1 lepers potintial pet formance ex: Snel proceuers (1oeeute 10 yun) multi thyeading® . . a oh. Muth threading 1 fatuve whicr enabia emuitiel’ dhveads to “head exeuk an a Slagle pyocenor in an ovelapelng moro)” wead 1) feu atomicuit of prouy Man jveads rake o (v00M4 Cnatitnveading threading: dil) simu Woneousty,-lrvea dng: » ascale & Beyond: upu UKd to Manipuiak 3D qvopris, Fick developed by Nvionr 7 4 GIP Computing Multinudia ¢ images 1494-14 prowy multiple trveds simutnsly~ Modun gpu pyeuy twoay conwuwent -+nveady: Me aaa floating boint operations 4 data- intentve calculen a i Gna; > Fermi Gpu: improved mremory accey, genau cathe hievarny » shaves memory among stypagning - | 30 billion transistors , 512 lorey avvanged Onlbhit (ream muttiprouy G memony technology, storage technology , Wide area netwobing: Memory KCWMIOgY: 4g tovage 4echnolyy: Memon, ry for harddrive Oa vw hae have Increaxd from dbome to ah YG ae bums _ on morogt |" using 0 auiyeom period wide prea netwoveing? (wan) rombps to 1 Gbps” 4 Vivtual machin? Cena thplentngs £4 Suspentions ati) Provision: *cluyiey Qycnilechves iQ A Clie ©) @ ) anneal Gateway @Q_ SAW, LAN Nal neko \ (s ; fi fn Ethernet myrindt, Ent | Band eke ey) eo | | [2 /odevicy) jpiskaway] a ee; ca, EE ORE Es ee SCouwity sy Crug . patwmonce metvia: OGY metrias mebig senna throughput @ system avariawility: se ot! Hrougnput a procuoy cpeed, riehwore pandwid th, respon muvig ousodatd Mo-of tale Cony out bya iyikm jn dpedee UA TEU mig, C million irutruln eer tccord) Dimensions OF (La 1a. bili ty Go Vie peotebiety? “iy fottwave ability” ciiy Application sccuablitty iv, Technology sealarvi lity: rete SE we Fault tolerances kyitm owollobithy: tereqs (1A Chighdyailbly) VU desived inal clarke, grias ,Papre A bys iF highly available 111 has long mean tre to faitare (MITE) & Shove mean time to Repaiv( SMTTE&)* Syuem avai (ouhility = MITE/ (MT TE + MTT eR) pvokction s unpaid Client, Hleqal shave Wopgrightt: Cory vight dive to Sytem % Networl: qlefesorie thvea ty integrity Loss oF Loss oF Loss of improper confidential ty unsgvity vat latbi lity AuthentiaHon [in - 7 [eae] “Fllegitimaue | eavage |= Lie A714 ae t _ x jtradts anlys4s -em| RE isan l masquerade Ve Pinete) } ¢ ves Avo peng, | Lurovivte | s poking aemeedt| | rapa aco aa Resour echauition| [Reradiation | |. weely | — [etree ted C computing= Aimry perfmnce goal in tonventional Energy ef fien ine clistribut systems ave high petformante % high throughout igtvibu kd computing § tion oF unued Sowers= uctul Y fore ae not ut hel » qh proey can rr Active (owed = pavallel % A Energy wonsawiP Seqvou Some ave unuctul Swed identthy y eliminate edu time | Spa:x i) Plementagi f plementation level oF Virtualization? CiHova) iv tualization } P virtualization is a procey that allows a computer fe snaik oH hlw Velouray With mubtiple digitally separaka environment UF if Mot early 40 implement ty WOSHUCHOn Ser Archiketuve level CISAJ : di HOVdwave AbstvacHon (evel ane dy operating system level (OSL)# : YD Ubvary Level (LU: aptifiom tibvany CW Application evel (AL): \ we Virtual eration Stucoalitsfratanin: — [mune] J | (exam Host ofevating 4m Hardware oun so (et ree | | vail coma we ra hn fave ince (or a} | ie | me [ater nee Hw | niw_| Hie 5 Viviwaliza tion | ayer Ley) ) Has host OF Kordwet vunnlg poor ool ol rae [voy 4 ae rordwore He vvtuativalon (ie = \Av Anattization Avchi tecture ¥ Kypavisors tis lio now 04 Virtual ma caine moron: ml | te petdovey vivtualation opesottiony [ravi Virualina ion cy Naive Hyfuvisore 2 Sty Hosted Hypewviror ppp ANP AED mer Mueaviade) sonlee quel os — (uci os einicrasofe Hypory Hypoivisoy/viams ’ * Vinwante © opfcLt Virlual Ex * Kernal virual machine (KYM) y Workstation pro 1,2