Resource Leasing Cloud Computing Model: A Win-Win Strategy for Resource Owners and Cloud Service Providers
Content
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
252
Resource Leasing Cloud Computing Model: A Win-Win Strategy for Resource Owners and Cloud Service Providers
John Thomas, Kalyana Raman, Vijay K. Chaurasiya, Santanu Das
Abstract—Investment on computing resources are pricey for any business, which holds the capacity to serve their clients require extensive resource outfit such as CPU, Storage, Memory, Software etc. In this context cloud computing is a paradigm which resolved the upfront cost on resources by its elastically scalable, economically saleable and globally located resources called cloud without investing more into hardware and software. The problem vested not in availing the resources through either cloud or the other; instead it vests in utilising the acquired computing resources optimally without being kept it idle. In this paper we analyse the serious issue of wastage of computing resources of an individual user or organization during the idle hours and propose a model for leasing their computing resource to cloud service provider for better utilisation of the same. In this resource leasing cloud model the owner of the computing resource lease his computing resource during the idle hours to the cloud service provider. It will enhance the resource utilization of the organization by augmenting the resource capacity of cloud service provider with minimal cloud upfront cost incurring by the cloud service provider. Also the model proposed an incentive scheme which will encourage organizations to participate in the process to augment the resource capacity of cloud service providers. Index Terms—Distributed Architectures, Emerging Technologies
—————————— ——————————
1 INTRODUCTION
A
ccording to NIST (National Institute of Standards and Technology), cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction [1]. With the rapid change in technology, replacement of computing resources such as Processor, Storage, Memory, Software etc., are unavoidable. This brings forth certain overhead for an organization on investment towards new hardware and software, along with dumping the existing hardware and software that are functioning properly. Resolving these overhead by any means may render greater impacts to any organization on their business. Gartner forecasts the worldwide Information Technology (IT) spending to total $3.8 trillion for 2012 (a 3.7 percent increase from 2011) and the spending on computing hardware alone to reach $424 Billion (a 5.1 percent increase from 2011) [2]. Generally cost is the most important factor for any business to deliver contemporary IT solutions. It may seems like IT can resolve the business overhead through its delivery but a significant amount of investment is required for acquirement and advancement of the same would of course captive savings of any business. In that context cloud computing renders on-demand IT infrastructure that lets business to consume exactly the amount of resources
————————————————
they actually need. Further businesses are not limited to fix their consumption of the storage, bandwidth or computing resources because it is elastically scalable. But being provided that sophistication of elastic scalability, it mandates businesses to provision their necessary and maximise their ROI (Return on Investment) instead of acquiring too many computing resources and wasting the same. Hence we require a new model that enables maximum utilization of the computing resources.
2 INFRASTRUCTURE SHARING MODEL FOR
MAXIMUM VALUE
After more than establishing an environment for computing either by corporate or home user, it will become essential for them to effectively utilise the established resources optimally according to their need parallel to periodic upgrade of the same. When it comes to optimum utilization of resources, the computing resources that are idle should be taken into note and also the cost for upgrading it also to be taken into consideration.
2.1 Idle time of computers is a waste of resources
Investment on computers is capital expenditure for any business which adds value out of its resources and is subjective to cost periodically. Besides utilising the resources of a computer for business, it is pertinent that such costly gadget ever not to be kept idle. The resources of any computer include the components such as processor, memory etc., the unused processing power is a wastage and loss for the business when reckoning the value and we should not fail to deduct with the depreciation value
• John Thomas, Kalyana Raman, Dr. Vijay K. Chaurasiya, Dr. Santanu Das Authors are with MBA & MS-CLIS Division, Indian Institute of Information Technology, Allahabad, India
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
253
of the same computing resources. Hence through this paper we propose a model that enables users to decide on leasing out their exiting computing resources during its idle time. We believe that it is possible to export, apply and adapt this philosophy to the Cloud computing paradigm. By involving users to this leasing towards Cloud will promote utilization of idle computing resources in cloud whereas the computing power of cloud will rose to maximum. This new computing paradigm gives the power and the control to the resource owners, who can decide how to maximize their resource utilisation efficiently through globally and geographically located clouds. The resource owners can also earn money through this concept by leasing their resources to Cloud computing according to their resources capabilities. Therefore, in this resource leasing cloud model both the commercial and non-commercial viewpoints coexist: in the former case the resource-owner extends orientation towards Cloud keeping the idle time of resources as a means to earn whereas in latter case resource owner will be optimizing the idle time of computing resources.
lease its high end computing resource to a user of Asian continent through a cloud service provider during idle hours, where the night time of North America would be day time in Asia. This resolves the unreasonable incurrence of expenditure by any corporate or home user towards computing resource establishment and maintenance which is eventually evolving with certain period of time and place.
2.5 Other Approaches
Earlier proposals and work towards utilizing the free computing resources such as CPU cycles and storage through Grid Computing [5] included the volunteer computing model where the owners of the computer volunteered their idle resource time through grid utility to research and distributed scientific computing [6]. TeraGrid [7] and Open Science Grid [8]were some of the remarkable pro-jects in Grid-based resource sharing. Adaption of this volunteer Computing to the cloud, the Cloud@Home model, was an enhancement of the grid-utility vision of Cloud computing [9]. Both volunteer Computing in grid and Cloud@Home were based on the free CPU cycles available during the time of owner using his computer. While both these models are suitable for distributed scientific computing and batch-type jobs, the drawback is that it does not allow the prediction of available resources at any time and may not be suitable for any time constrained or performance constrained or regular computing uses.
2.2 Infrastructure upgrades are inevitable
While governing the total IT infrastructure of any business, the hardware and software that are deployed onto business environment should be monitored. Because the business itself being dependent on infrastructure, therefore both hardware and software and upgrades, if needed any, for the functioning of the business environment should be addressed periodically. The reason behind this philosophy is that the technology is moving in pace where business is evolving through that. Hence when there is an upgrade of hardware or software in market or there is a trend in market that certain software/hardware are essential to meet the demand, then there comes an inevitable circumstance where the business ought to upgrade their software and hardware.
3 RESOURCE LEASING CLOUD MODEL
We propose, a “resource leasing cloud model” that addresses the drawbacks discussed in section 2. We use the NIST (National Institute of Standards and Technology) definitions, cloud subscriber or subscriber for person or organization that is a customer of a cloud; and cloud provider or provider for an organization that provides cloud services [1]. We also define a new term cloud lessor or lessor as a person or organization who leases his computing resources to the cloud provider. As an open source cloud implementation and considering the availability, we use the Eucalyptus/Ubuntu Enterprise Cloud (UEC) architecture [4] for explaining the concept. We specifically use the following terms from the Eucalyptus/UEC vocabulary. • Cloud Controller (CLC): Managing and exposing the underlying virtualized resources (machines (servers), network, and storage) via user-facing APIs is the responsibility of the Cloud Controller. • Cluster Controller (CC): The CC controls the execution of virtual machines (VMs) running on the nodes and manages the virtual networking between VMs and between VMs and external users. • Node Controller (NC): Through the functionality of a hypervisor, the Node controller controls VM activities, including the execution, inspection, and termination of VM instances. • Storage Controller (SC): The SC provides blocklevel network storage that can be dynamically attached
2.3 Expenditure on Cloud Infrastructure
In cloud computing the Subscriber need not spend money on acquiring computing infrastructure, but it is a shift of capital expenditure from the Subscriber to the Cloud Provider. IDC (International Data Corporation) says that overall spending by public cloud service providers on storage hardware, software, and professional services will grow at a compound annual growth rate (CAGR) of 23.6% from 2010 to 2015, while enterprise spending on storage for the private cloud will experience a CAGR of 28.9%. By 2015, combined spending for public and private cloud storage will be $22.6 billion worldwide [3]. Hence any model that can reduce the spending on cloud infrastructure will become viable solution for the business.
2.4 Time-Zone difference – A great possibility
Resource holders of different time zones being located across different continents provide them a great advantage to share their computing resources during its idle hours. Computing resource holder of North America can
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
254
by VM [4] Ms. Co onsidering that majority of personal com f mputers, especia ally in offices and in software developm ment centres, are id during nig time and the users locat in differdle ght ted ent ti ime zones only require com mputing resou urces at differen period of tim that are not overlappin with each nt me n ng other enables the design of a res r d source sharing model that g can b in place to share the computing resour be s rces between those users to fulfil their comput e l ting needs. In the proposed resource leas n d sing cloud mo odel, the lessor le eases the comp puting resources to the clou provider, ud when his computin resources are idle, throug cloud for n ng a gh a par rticular period of time. Th must be differentiated d his from the grid comp puting [5]. In grid computin the ownng ng s er of the computin resource and the grid share the resourc simultaneo ces ously and the grid is taking advantage g of the free resourc time. With grid computi e ce ing, you can provi ision computi ing resources as a utility that can be turne on or off [1 whereas in our propos resource ed 10], i sed leasin cloud mod ng del, during th lease time, the leased he , comp puting resourc is not used by the lessor and is comce pletel available to the cloud pro ly o ovider.
vailable for le ease from a s single numbers of computers av ajority of the computers av vailable for leasing location, ma can turn int Node Cont to trollers, and f few dedicated mad chines can a as Cluster Controller an Storage Con act nd ntroller. This wil allow all the machines in the leasing fa ll e acility form a ‘cluste in the clou Figure 2 sh er’ ud. hows together to f the logical re epresentation of Corporate Leasing Mode el.
Fig. 2. Cor rporate Leasing M Model
3.1.1 Use C Case for Corp porate Leasing Model g
A software firm in New Jersey, USA has 1200 per rsonal computers w with high end configuration Its normal w n. working hours ar from 9 AM to 5 PM EST. After the working re hours aroun 1000 comp nd puters are not in use. The New t Jersey firm d decided to lea the 1000 m ase machines from 6 PM to 8AM thro ough cloud an made a con nd ntract with a c cloud provider. One Softw tware develop pment firm in Bangalore, India n require 1000 high end com 0 mputers, for a time period o one of roject. They n year, for an outsourced pr normally work durk ing 9 AM to 5 PM IST. C o Currently the have 1000 spare ey machines w which are 6 ye ears old and not suited fo the or developmen work of th new projec Bangalore firm nt he ct. decided to le ease high end computers fr d rom cloud for their developmen work and a nt access them th hrough the old mad chines availa able with them m. In the ab bove scenario the cloud provider can ge the et resources fro the New Jersey firm on lease agreem n ment, om saving a lot on infrastruct ture investme the New J ent; Jersey firm can earn money thro rn ough leasing, m making more v value from their in investment on infrastructur and the Ba n re; angalore firm can take compu n uters on lease for their dev velopment saving the money f acquiring n g for new hardware All e. these happe through clo en oud and the c cloud provider acts r as an agent b between the tw firms. wo
Fig. 1. Logical Separation of the system between lessor and provider m r
Se eparating the computing re esources for own use by o the le essor and for leasing to the provider du e uring its idle hours can be confi s igured in a machine as a dual booting m d mech hanism. Figure 1 shows a general view of dual boote o ing m mechanism for the resource leasing cloud model. One r partit tion boots to the lessor’s ow Operating System and t wn deskt whereas th other partition boots to a cloud suptop he portin Operating System like Ubuntu Enterp ng U prise Cloud. The l lessor restarts the machine after his own use manuala u ly or through sche eduling which will automa h atically boot the o other Operatin System an the machi ng nd ine registers itself to the cloud with the preconfigured settin w ngs.
3.1 C Corporate Le easing Model
In Co orporate world a single com d, mpany will hav hundreds ve to tho ousands or more personal computers. De c epending on the b business type and working time, many of those coma t f puter will be free or switched off for many hours, espers h cially during non-w y working hours Considering those large s. g
3.2 Home L Leasing Mod del
Many home users use th e heir desktop computers at h home for minimal hours. In m l most cases dur ring day time and e
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
255
office h hours the com mputers will be free or switc e ched off. In develo oped cities wit high bandw th width infrastru ucture and high p population th hese computer together ca form a rs an cluster and join a pa of a cloud. If the cloud pr r art I rovider can setup Cluster Contr roller and Sto orage Control ller in that y, me a locality all the hom computers that are all available to lease c turn into Node Controll can N lers and form part of the cloud. Figure 3 show the logical representation of Home ws ng Leasin Model.
3.3.1 Role of Service Leve Agreement f vel t
As the resour leasing clo rce oud model in nvolves one m more party, the Le essor, into the picture the relationship bee e tween the Le essor and the Cloud Provi e ider will beco ome more comple In order t avoid intricacies the rig ex. to ghts, duties, respon nsibilities, ter rms and cond ditions should be d clearly stated in the SLA (S Service Level Agreement). T This includes term and conditi ms ions for ensur ring the availa ability of the com mputing resou urces during th agreed leas he sing time, avoiding intrusion att g tempts to the other partition of the leased com mputer etc. SL should be clear and un LA e nambiguous in ev very aspect so that it can r o resolve the iss sues when the trus between the parties is bro st e oken or when the n technical cont trols fail.
3.3.2 Segreg gating resourc ces
In leasing mo odel every com mputer is used by two parties; one is the own of the com ner mputer, the Le essor, and the other is the clou subscriber. So all the so ud . oftware and d data needs to be se egregated and secured from each other. T d m Two separate wor rking environ nments, indep pendent of e each other and sec cure from acci idental or mal licious use are ree quired. We used t dual booti the ing mechanism for our exp m periments. This is enough for t segregatio but may no be s the on ot completely se ecure as it is, b customize versions of opbut ed erating system with secu ms urity restrictions can prov vide more security Further the cloud opera y. e ating systems extending suppo to this mod can be dev ort del veloped in futu ure. Type 1 hype ervisor (or n native, bare-m metal) is another mechanism h having potent tial to segrega user work ate king environments without muc performanc degradation as s ch ce n long as all the system resou e urces are utiliz for runnin a zed ng single image at a time. We tested the N e NxTop Engine [11] ypervisor, one which allows desktop virt e s tualbare-metal hy ization, howe ever we did no use it for ou experiment as ot ur ts it need furthe refinements to suit our pu er urpose.
Fig. 3. Home Leasing Model 3 g
3.2.1 Use Case for Home Leasin Model r ng
A Clin nical Research Company in New Jersey, USA needs U high v volume of bac ckground com mputing durin their ofng fice ho ours from 9 AM to 5 PM ES They don’t need any ST. frequent front end interaction for their application; it can i r run as automated background work. They hav decided s b w ve to lease the compute from a clou provider. ers ud ng opulated area with lots of home comh Bein a highly po puters, the cloud pr rovider decide to take com ed mputers on lease f from home users who are not using th heir system from 9 AM to 5 PM EST for pro M oviding it to th Clinical he Resear Company. Cloud provider setup a cl rch . loud infrastructu in that lo ure ocality with Cluster Contr C rollers and Storag Controllers and the lease home com ge ed mputers are made N Node Controllers. In t above scen the nario, the clou provider acquires reud a source from local home users and providing it to the es a g Clinica Research Company thro al C ough cloud. Where the W cloud provider and the Clinical Research Com R mpany save y u ructure, the home users h money on setting up the infrastr are ma aking money by leasing th heir computer during its idle ho ours and addin value to the investment ng eir t.
3.3.3 Controll of Servers
Though the r relations betw ween various parties are g governed by SLA to avoid more complexity and data lock A, y k-in, vers can be ve control of serv ested with the Cloud Provi e ider. In our Home Leasing mo e odel, Cluster Controllers and Storage Contr rollers are esta ablished by th cloud Provider he itself wherea s in Corpora Leasing m ate model the Clu uster nd ontroller are at Lessor premi t ises. Controllers an Storage Co These servers can be establ s lished by the C Cloud Provide at er the Lessor pre emises and co ontrolled by th Provider itself. he However resp ponsibility of providing ph hysical security to y the servers th are deplo hat oyed at the Le essor premise is es vested with t lessor, wh the here the stipul lations and as ssertions towards physical secu s urity measure can be attai es ined through the S SLA. The conc cerns and deb bate on the ph hysical security of hypervisors that exists in cloud computing f will remain in this model w n with an added concern that the d t hypervisor is residing on th lessor prem he mises and the lessor is having physical acces to it. Any a ss attempt or mec chanism to add dress such sec curity issues and auditability should obviou usly be extend to include the lessor also ded e o.
3.3 Iss sues in the Leasing Mod L del
We are not addressi the issues that already exist in the e ing e cloud computing pa aradigm, but only the issues pertaining to t new mode proposed. the el
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
/
256
4 PROOF OF CONCEPT O
We u used Eucalyp ptus/Ubuntu Enterprise Cl loud (UEC), Ubun Server 10.10 [12], for se ntu etting up resource leasing cloud environment in a University LAN consi d t isting multiple s subnets. We followed the Eucalyptus Beginner's e s Guid v2.0 [4] to se de etup the cloud environment d t. Th installation and configu he n uration instruc ctions given in the Eucalyptus guide were fo e g ollowed exactly with Servy er 1 h having the Clo Controller Cluster Con oud r, ntroller, Walrus a and Storage Controller. Bu both netwo interface C ut ork cards of Server1 were connected to the public switch so s w that t Node Con the ntrollers (Lesso machines) that needs to or t join t cloud by co the onnecting to th Cluster Con he ntroller. In nstead of havin a single No Controller as Server 2 ng ode r we ha installed and configured multiple No Controlave a ode lers a in Figure 4. All the Node Controllers were installed as on co omputers hav other boards supporting ving Intel mo virtua alization tech rs hnology. All the computer acting as Node Controllers were made du booting wit one partie w th ual tion having Wind nd dows Operatin System an the other ng partit tion is having Ubuntu Server 10.10 working as Node Contr rollers.
ugh plained in this paper enable the s es model throu cloud, exp d source harnes ssing and opti imize unused and untapped res efficiency of cl loud through this exporting cong further the e f cept, which diminish the fading-out of costly computing CPU cycles, P Processor, mem mory) due to idleresources (C n ower of cloud pavd ness and can escalate the computing po ners to lease a and earn. Prov vided ing way for resource own d be ly ability the proposed model will b feasible onl if the availa andwidth is co onsidof bandwidt is high and the cost of ba th ost puting resourc ces. erably cheap than the co of the comp per ch er costly The optim mistic approac of this pape is that the c uld le, hould computing r resources shou not be idl if idle it sh the lizing be utilized s somehow or t other. However by util r the same the should be some benefit for the lessor who ere ready to share the obviously p possess the va aluable and r n e same. This m makes a win-win situation for both the proopt odolovider and th lessor. Ther he refore we ado our metho mmodate resources which can dra gy to cloud w astically accom he e across globe and periodic cally release th same to see ekers. stage of computing that idle comp We believe t puters are was ell the power and t proposed model can we suffice the existombining with the ing scarcity of computing resources co g cost effective cloud. e
REFERENC ES
National Institute of Standards & Technology (NIST), http://csrc rc.nist.gov/publica ations/drafts/800-1 146/Draft-NIST-SP P800146.pdf, 20 011 /www.gartner.com m/it/page.jsp?id=7 707508, [2] Gartner N Newsroom, http:// 2012 [3] IDC Press R Release, http://ww oc.jsp?containerId= =prUS23097611, 20 011 ww.idc.com/getdo [4] Eucalyptu Guide U UEC Edition v2.0, us Beginner's http://csso ss.com/2010/12/eu ucabookv2-0.pdf, 2 2010 soss.files.wordpress [5] Wikipedia http://en.wikipe edia.org/wiki/Gri id_computing, 201 12 a, [6] David P. A les omputational and s storage Anderson and Gill Fedak, "The co potential o volunteer comp puting," In CCGR ’06, pp. 73–80 IEEE RID 0, of Computer Society, 2006. r [7] Catlett, C., Beckman, P., Sko D., and Foster, I., “Creating and operatow, , rly, ing nationa ”, structure services.” CTWatch Quarter vol. nalscale cyberinfras 2, no. 2, pp 2-10, 2006. p. grid.org, 2011 [8] Open Scien Grid, http://w www.openscienceg nce [9] Vincenzo D. Cunsolo, Salva Antonio Puliafito, Marco vatore Distefano, A @Home Scarpa, "Vo Cloud: The Cloud@ ing Volunteer Computi and Desktop C rk mposium on Networ ComParadigm, Proc. Eighth IEEE International Sym E ," 9.41. 10.1109/NCA.2009 puting and Applications, pp.13 34-139, 2009, DOI 1 ve computing computing IBM, [10] Cloud ersus grid c http://ww ww.ibm.com/deve b/library/wa-cloud dgrid/, eloperworks/web 2012 /nxtopputer.com/nxtop/ [11] NxTop Engine, http:// /www.virtualcomp workstatio 2012 on, [12] Ubuntu 10 (Maverick Mee 0.10 /10.10/, ases.ubuntu.com/ erkat), http://relea 2012 [1]
Fig. 4. Test Setup
Te setup in th LAN was similar to the Home Leasest he Model proposed above. Bu we did not measure or ing M ut pare the perfo comp ormance as th performance in a real he uction implem produ mentation will vary from th test setup he l high speed LA in a h AN.
5 CONCLUSION
dical escalation of cost for acquiring ren a To su urmise the rad e h sourc and post acquirement expenses such as mainteces a esources put wards those re nance upgrades, patches etc. tow e, p e resou urce holder a standpoint thereof. The inevitable sting or alread established computing d dy chang to the exis ges pgrade are also not static for a longer f envir ronment as up hereof the com rces that are mputing resour perio of time. Th od tive of time period will no justify the p ot kept idle irrespect nses that are ost reaso onableness of the overall co and expen ncurring dayny -by-day by an of the reall in ncurred and in ence the conc urce leasing cept of resou sourc owner. He ce
252
Resource Leasing Cloud Computing Model: A Win-Win Strategy for Resource Owners and Cloud Service Providers
John Thomas, Kalyana Raman, Vijay K. Chaurasiya, Santanu Das
Abstract—Investment on computing resources are pricey for any business, which holds the capacity to serve their clients require extensive resource outfit such as CPU, Storage, Memory, Software etc. In this context cloud computing is a paradigm which resolved the upfront cost on resources by its elastically scalable, economically saleable and globally located resources called cloud without investing more into hardware and software. The problem vested not in availing the resources through either cloud or the other; instead it vests in utilising the acquired computing resources optimally without being kept it idle. In this paper we analyse the serious issue of wastage of computing resources of an individual user or organization during the idle hours and propose a model for leasing their computing resource to cloud service provider for better utilisation of the same. In this resource leasing cloud model the owner of the computing resource lease his computing resource during the idle hours to the cloud service provider. It will enhance the resource utilization of the organization by augmenting the resource capacity of cloud service provider with minimal cloud upfront cost incurring by the cloud service provider. Also the model proposed an incentive scheme which will encourage organizations to participate in the process to augment the resource capacity of cloud service providers. Index Terms—Distributed Architectures, Emerging Technologies
—————————— ——————————
1 INTRODUCTION
A
ccording to NIST (National Institute of Standards and Technology), cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction [1]. With the rapid change in technology, replacement of computing resources such as Processor, Storage, Memory, Software etc., are unavoidable. This brings forth certain overhead for an organization on investment towards new hardware and software, along with dumping the existing hardware and software that are functioning properly. Resolving these overhead by any means may render greater impacts to any organization on their business. Gartner forecasts the worldwide Information Technology (IT) spending to total $3.8 trillion for 2012 (a 3.7 percent increase from 2011) and the spending on computing hardware alone to reach $424 Billion (a 5.1 percent increase from 2011) [2]. Generally cost is the most important factor for any business to deliver contemporary IT solutions. It may seems like IT can resolve the business overhead through its delivery but a significant amount of investment is required for acquirement and advancement of the same would of course captive savings of any business. In that context cloud computing renders on-demand IT infrastructure that lets business to consume exactly the amount of resources
————————————————
they actually need. Further businesses are not limited to fix their consumption of the storage, bandwidth or computing resources because it is elastically scalable. But being provided that sophistication of elastic scalability, it mandates businesses to provision their necessary and maximise their ROI (Return on Investment) instead of acquiring too many computing resources and wasting the same. Hence we require a new model that enables maximum utilization of the computing resources.
2 INFRASTRUCTURE SHARING MODEL FOR
MAXIMUM VALUE
After more than establishing an environment for computing either by corporate or home user, it will become essential for them to effectively utilise the established resources optimally according to their need parallel to periodic upgrade of the same. When it comes to optimum utilization of resources, the computing resources that are idle should be taken into note and also the cost for upgrading it also to be taken into consideration.
2.1 Idle time of computers is a waste of resources
Investment on computers is capital expenditure for any business which adds value out of its resources and is subjective to cost periodically. Besides utilising the resources of a computer for business, it is pertinent that such costly gadget ever not to be kept idle. The resources of any computer include the components such as processor, memory etc., the unused processing power is a wastage and loss for the business when reckoning the value and we should not fail to deduct with the depreciation value
• John Thomas, Kalyana Raman, Dr. Vijay K. Chaurasiya, Dr. Santanu Das Authors are with MBA & MS-CLIS Division, Indian Institute of Information Technology, Allahabad, India
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
253
of the same computing resources. Hence through this paper we propose a model that enables users to decide on leasing out their exiting computing resources during its idle time. We believe that it is possible to export, apply and adapt this philosophy to the Cloud computing paradigm. By involving users to this leasing towards Cloud will promote utilization of idle computing resources in cloud whereas the computing power of cloud will rose to maximum. This new computing paradigm gives the power and the control to the resource owners, who can decide how to maximize their resource utilisation efficiently through globally and geographically located clouds. The resource owners can also earn money through this concept by leasing their resources to Cloud computing according to their resources capabilities. Therefore, in this resource leasing cloud model both the commercial and non-commercial viewpoints coexist: in the former case the resource-owner extends orientation towards Cloud keeping the idle time of resources as a means to earn whereas in latter case resource owner will be optimizing the idle time of computing resources.
lease its high end computing resource to a user of Asian continent through a cloud service provider during idle hours, where the night time of North America would be day time in Asia. This resolves the unreasonable incurrence of expenditure by any corporate or home user towards computing resource establishment and maintenance which is eventually evolving with certain period of time and place.
2.5 Other Approaches
Earlier proposals and work towards utilizing the free computing resources such as CPU cycles and storage through Grid Computing [5] included the volunteer computing model where the owners of the computer volunteered their idle resource time through grid utility to research and distributed scientific computing [6]. TeraGrid [7] and Open Science Grid [8]were some of the remarkable pro-jects in Grid-based resource sharing. Adaption of this volunteer Computing to the cloud, the Cloud@Home model, was an enhancement of the grid-utility vision of Cloud computing [9]. Both volunteer Computing in grid and Cloud@Home were based on the free CPU cycles available during the time of owner using his computer. While both these models are suitable for distributed scientific computing and batch-type jobs, the drawback is that it does not allow the prediction of available resources at any time and may not be suitable for any time constrained or performance constrained or regular computing uses.
2.2 Infrastructure upgrades are inevitable
While governing the total IT infrastructure of any business, the hardware and software that are deployed onto business environment should be monitored. Because the business itself being dependent on infrastructure, therefore both hardware and software and upgrades, if needed any, for the functioning of the business environment should be addressed periodically. The reason behind this philosophy is that the technology is moving in pace where business is evolving through that. Hence when there is an upgrade of hardware or software in market or there is a trend in market that certain software/hardware are essential to meet the demand, then there comes an inevitable circumstance where the business ought to upgrade their software and hardware.
3 RESOURCE LEASING CLOUD MODEL
We propose, a “resource leasing cloud model” that addresses the drawbacks discussed in section 2. We use the NIST (National Institute of Standards and Technology) definitions, cloud subscriber or subscriber for person or organization that is a customer of a cloud; and cloud provider or provider for an organization that provides cloud services [1]. We also define a new term cloud lessor or lessor as a person or organization who leases his computing resources to the cloud provider. As an open source cloud implementation and considering the availability, we use the Eucalyptus/Ubuntu Enterprise Cloud (UEC) architecture [4] for explaining the concept. We specifically use the following terms from the Eucalyptus/UEC vocabulary. • Cloud Controller (CLC): Managing and exposing the underlying virtualized resources (machines (servers), network, and storage) via user-facing APIs is the responsibility of the Cloud Controller. • Cluster Controller (CC): The CC controls the execution of virtual machines (VMs) running on the nodes and manages the virtual networking between VMs and between VMs and external users. • Node Controller (NC): Through the functionality of a hypervisor, the Node controller controls VM activities, including the execution, inspection, and termination of VM instances. • Storage Controller (SC): The SC provides blocklevel network storage that can be dynamically attached
2.3 Expenditure on Cloud Infrastructure
In cloud computing the Subscriber need not spend money on acquiring computing infrastructure, but it is a shift of capital expenditure from the Subscriber to the Cloud Provider. IDC (International Data Corporation) says that overall spending by public cloud service providers on storage hardware, software, and professional services will grow at a compound annual growth rate (CAGR) of 23.6% from 2010 to 2015, while enterprise spending on storage for the private cloud will experience a CAGR of 28.9%. By 2015, combined spending for public and private cloud storage will be $22.6 billion worldwide [3]. Hence any model that can reduce the spending on cloud infrastructure will become viable solution for the business.
2.4 Time-Zone difference – A great possibility
Resource holders of different time zones being located across different continents provide them a great advantage to share their computing resources during its idle hours. Computing resource holder of North America can
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
254
by VM [4] Ms. Co onsidering that majority of personal com f mputers, especia ally in offices and in software developm ment centres, are id during nig time and the users locat in differdle ght ted ent ti ime zones only require com mputing resou urces at differen period of tim that are not overlappin with each nt me n ng other enables the design of a res r d source sharing model that g can b in place to share the computing resour be s rces between those users to fulfil their comput e l ting needs. In the proposed resource leas n d sing cloud mo odel, the lessor le eases the comp puting resources to the clou provider, ud when his computin resources are idle, throug cloud for n ng a gh a par rticular period of time. Th must be differentiated d his from the grid comp puting [5]. In grid computin the ownng ng s er of the computin resource and the grid share the resourc simultaneo ces ously and the grid is taking advantage g of the free resourc time. With grid computi e ce ing, you can provi ision computi ing resources as a utility that can be turne on or off [1 whereas in our propos resource ed 10], i sed leasin cloud mod ng del, during th lease time, the leased he , comp puting resourc is not used by the lessor and is comce pletel available to the cloud pro ly o ovider.
vailable for le ease from a s single numbers of computers av ajority of the computers av vailable for leasing location, ma can turn int Node Cont to trollers, and f few dedicated mad chines can a as Cluster Controller an Storage Con act nd ntroller. This wil allow all the machines in the leasing fa ll e acility form a ‘cluste in the clou Figure 2 sh er’ ud. hows together to f the logical re epresentation of Corporate Leasing Mode el.
Fig. 2. Cor rporate Leasing M Model
3.1.1 Use C Case for Corp porate Leasing Model g
A software firm in New Jersey, USA has 1200 per rsonal computers w with high end configuration Its normal w n. working hours ar from 9 AM to 5 PM EST. After the working re hours aroun 1000 comp nd puters are not in use. The New t Jersey firm d decided to lea the 1000 m ase machines from 6 PM to 8AM thro ough cloud an made a con nd ntract with a c cloud provider. One Softw tware develop pment firm in Bangalore, India n require 1000 high end com 0 mputers, for a time period o one of roject. They n year, for an outsourced pr normally work durk ing 9 AM to 5 PM IST. C o Currently the have 1000 spare ey machines w which are 6 ye ears old and not suited fo the or developmen work of th new projec Bangalore firm nt he ct. decided to le ease high end computers fr d rom cloud for their developmen work and a nt access them th hrough the old mad chines availa able with them m. In the ab bove scenario the cloud provider can ge the et resources fro the New Jersey firm on lease agreem n ment, om saving a lot on infrastruct ture investme the New J ent; Jersey firm can earn money thro rn ough leasing, m making more v value from their in investment on infrastructur and the Ba n re; angalore firm can take compu n uters on lease for their dev velopment saving the money f acquiring n g for new hardware All e. these happe through clo en oud and the c cloud provider acts r as an agent b between the tw firms. wo
Fig. 1. Logical Separation of the system between lessor and provider m r
Se eparating the computing re esources for own use by o the le essor and for leasing to the provider du e uring its idle hours can be confi s igured in a machine as a dual booting m d mech hanism. Figure 1 shows a general view of dual boote o ing m mechanism for the resource leasing cloud model. One r partit tion boots to the lessor’s ow Operating System and t wn deskt whereas th other partition boots to a cloud suptop he portin Operating System like Ubuntu Enterp ng U prise Cloud. The l lessor restarts the machine after his own use manuala u ly or through sche eduling which will automa h atically boot the o other Operatin System an the machi ng nd ine registers itself to the cloud with the preconfigured settin w ngs.
3.1 C Corporate Le easing Model
In Co orporate world a single com d, mpany will hav hundreds ve to tho ousands or more personal computers. De c epending on the b business type and working time, many of those coma t f puter will be free or switched off for many hours, espers h cially during non-w y working hours Considering those large s. g
3.2 Home L Leasing Mod del
Many home users use th e heir desktop computers at h home for minimal hours. In m l most cases dur ring day time and e
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
255
office h hours the com mputers will be free or switc e ched off. In develo oped cities wit high bandw th width infrastru ucture and high p population th hese computer together ca form a rs an cluster and join a pa of a cloud. If the cloud pr r art I rovider can setup Cluster Contr roller and Sto orage Control ller in that y, me a locality all the hom computers that are all available to lease c turn into Node Controll can N lers and form part of the cloud. Figure 3 show the logical representation of Home ws ng Leasin Model.
3.3.1 Role of Service Leve Agreement f vel t
As the resour leasing clo rce oud model in nvolves one m more party, the Le essor, into the picture the relationship bee e tween the Le essor and the Cloud Provi e ider will beco ome more comple In order t avoid intricacies the rig ex. to ghts, duties, respon nsibilities, ter rms and cond ditions should be d clearly stated in the SLA (S Service Level Agreement). T This includes term and conditi ms ions for ensur ring the availa ability of the com mputing resou urces during th agreed leas he sing time, avoiding intrusion att g tempts to the other partition of the leased com mputer etc. SL should be clear and un LA e nambiguous in ev very aspect so that it can r o resolve the iss sues when the trus between the parties is bro st e oken or when the n technical cont trols fail.
3.3.2 Segreg gating resourc ces
In leasing mo odel every com mputer is used by two parties; one is the own of the com ner mputer, the Le essor, and the other is the clou subscriber. So all the so ud . oftware and d data needs to be se egregated and secured from each other. T d m Two separate wor rking environ nments, indep pendent of e each other and sec cure from acci idental or mal licious use are ree quired. We used t dual booti the ing mechanism for our exp m periments. This is enough for t segregatio but may no be s the on ot completely se ecure as it is, b customize versions of opbut ed erating system with secu ms urity restrictions can prov vide more security Further the cloud opera y. e ating systems extending suppo to this mod can be dev ort del veloped in futu ure. Type 1 hype ervisor (or n native, bare-m metal) is another mechanism h having potent tial to segrega user work ate king environments without muc performanc degradation as s ch ce n long as all the system resou e urces are utiliz for runnin a zed ng single image at a time. We tested the N e NxTop Engine [11] ypervisor, one which allows desktop virt e s tualbare-metal hy ization, howe ever we did no use it for ou experiment as ot ur ts it need furthe refinements to suit our pu er urpose.
Fig. 3. Home Leasing Model 3 g
3.2.1 Use Case for Home Leasin Model r ng
A Clin nical Research Company in New Jersey, USA needs U high v volume of bac ckground com mputing durin their ofng fice ho ours from 9 AM to 5 PM ES They don’t need any ST. frequent front end interaction for their application; it can i r run as automated background work. They hav decided s b w ve to lease the compute from a clou provider. ers ud ng opulated area with lots of home comh Bein a highly po puters, the cloud pr rovider decide to take com ed mputers on lease f from home users who are not using th heir system from 9 AM to 5 PM EST for pro M oviding it to th Clinical he Resear Company. Cloud provider setup a cl rch . loud infrastructu in that lo ure ocality with Cluster Contr C rollers and Storag Controllers and the lease home com ge ed mputers are made N Node Controllers. In t above scen the nario, the clou provider acquires reud a source from local home users and providing it to the es a g Clinica Research Company thro al C ough cloud. Where the W cloud provider and the Clinical Research Com R mpany save y u ructure, the home users h money on setting up the infrastr are ma aking money by leasing th heir computer during its idle ho ours and addin value to the investment ng eir t.
3.3.3 Controll of Servers
Though the r relations betw ween various parties are g governed by SLA to avoid more complexity and data lock A, y k-in, vers can be ve control of serv ested with the Cloud Provi e ider. In our Home Leasing mo e odel, Cluster Controllers and Storage Contr rollers are esta ablished by th cloud Provider he itself wherea s in Corpora Leasing m ate model the Clu uster nd ontroller are at Lessor premi t ises. Controllers an Storage Co These servers can be establ s lished by the C Cloud Provide at er the Lessor pre emises and co ontrolled by th Provider itself. he However resp ponsibility of providing ph hysical security to y the servers th are deplo hat oyed at the Le essor premise is es vested with t lessor, wh the here the stipul lations and as ssertions towards physical secu s urity measure can be attai es ined through the S SLA. The conc cerns and deb bate on the ph hysical security of hypervisors that exists in cloud computing f will remain in this model w n with an added concern that the d t hypervisor is residing on th lessor prem he mises and the lessor is having physical acces to it. Any a ss attempt or mec chanism to add dress such sec curity issues and auditability should obviou usly be extend to include the lessor also ded e o.
3.3 Iss sues in the Leasing Mod L del
We are not addressi the issues that already exist in the e ing e cloud computing pa aradigm, but only the issues pertaining to t new mode proposed. the el
JOURNAL OF COMPUTING, VOLUME 4, ISSUE 5, MAY 2012, ISSN 2151-9617 https://sites.google.com/site/journalofcomputing WWW.JOURNALOFCOMPUTING.ORG
/
256
4 PROOF OF CONCEPT O
We u used Eucalyp ptus/Ubuntu Enterprise Cl loud (UEC), Ubun Server 10.10 [12], for se ntu etting up resource leasing cloud environment in a University LAN consi d t isting multiple s subnets. We followed the Eucalyptus Beginner's e s Guid v2.0 [4] to se de etup the cloud environment d t. Th installation and configu he n uration instruc ctions given in the Eucalyptus guide were fo e g ollowed exactly with Servy er 1 h having the Clo Controller Cluster Con oud r, ntroller, Walrus a and Storage Controller. Bu both netwo interface C ut ork cards of Server1 were connected to the public switch so s w that t Node Con the ntrollers (Lesso machines) that needs to or t join t cloud by co the onnecting to th Cluster Con he ntroller. In nstead of havin a single No Controller as Server 2 ng ode r we ha installed and configured multiple No Controlave a ode lers a in Figure 4. All the Node Controllers were installed as on co omputers hav other boards supporting ving Intel mo virtua alization tech rs hnology. All the computer acting as Node Controllers were made du booting wit one partie w th ual tion having Wind nd dows Operatin System an the other ng partit tion is having Ubuntu Server 10.10 working as Node Contr rollers.
ugh plained in this paper enable the s es model throu cloud, exp d source harnes ssing and opti imize unused and untapped res efficiency of cl loud through this exporting cong further the e f cept, which diminish the fading-out of costly computing CPU cycles, P Processor, mem mory) due to idleresources (C n ower of cloud pavd ness and can escalate the computing po ners to lease a and earn. Prov vided ing way for resource own d be ly ability the proposed model will b feasible onl if the availa andwidth is co onsidof bandwidt is high and the cost of ba th ost puting resourc ces. erably cheap than the co of the comp per ch er costly The optim mistic approac of this pape is that the c uld le, hould computing r resources shou not be idl if idle it sh the lizing be utilized s somehow or t other. However by util r the same the should be some benefit for the lessor who ere ready to share the obviously p possess the va aluable and r n e same. This m makes a win-win situation for both the proopt odolovider and th lessor. Ther he refore we ado our metho mmodate resources which can dra gy to cloud w astically accom he e across globe and periodic cally release th same to see ekers. stage of computing that idle comp We believe t puters are was ell the power and t proposed model can we suffice the existombining with the ing scarcity of computing resources co g cost effective cloud. e
REFERENC ES
National Institute of Standards & Technology (NIST), http://csrc rc.nist.gov/publica ations/drafts/800-1 146/Draft-NIST-SP P800146.pdf, 20 011 /www.gartner.com m/it/page.jsp?id=7 707508, [2] Gartner N Newsroom, http:// 2012 [3] IDC Press R Release, http://ww oc.jsp?containerId= =prUS23097611, 20 011 ww.idc.com/getdo [4] Eucalyptu Guide U UEC Edition v2.0, us Beginner's http://csso ss.com/2010/12/eu ucabookv2-0.pdf, 2 2010 soss.files.wordpress [5] Wikipedia http://en.wikipe edia.org/wiki/Gri id_computing, 201 12 a, [6] David P. A les omputational and s storage Anderson and Gill Fedak, "The co potential o volunteer comp puting," In CCGR ’06, pp. 73–80 IEEE RID 0, of Computer Society, 2006. r [7] Catlett, C., Beckman, P., Sko D., and Foster, I., “Creating and operatow, , rly, ing nationa ”, structure services.” CTWatch Quarter vol. nalscale cyberinfras 2, no. 2, pp 2-10, 2006. p. grid.org, 2011 [8] Open Scien Grid, http://w www.openscienceg nce [9] Vincenzo D. Cunsolo, Salva Antonio Puliafito, Marco vatore Distefano, A @Home Scarpa, "Vo Cloud: The Cloud@ ing Volunteer Computi and Desktop C rk mposium on Networ ComParadigm, Proc. Eighth IEEE International Sym E ," 9.41. 10.1109/NCA.2009 puting and Applications, pp.13 34-139, 2009, DOI 1 ve computing computing IBM, [10] Cloud ersus grid c http://ww ww.ibm.com/deve b/library/wa-cloud dgrid/, eloperworks/web 2012 /nxtopputer.com/nxtop/ [11] NxTop Engine, http:// /www.virtualcomp workstatio 2012 on, [12] Ubuntu 10 (Maverick Mee 0.10 /10.10/, ases.ubuntu.com/ erkat), http://relea 2012 [1]
Fig. 4. Test Setup
Te setup in th LAN was similar to the Home Leasest he Model proposed above. Bu we did not measure or ing M ut pare the perfo comp ormance as th performance in a real he uction implem produ mentation will vary from th test setup he l high speed LA in a h AN.
5 CONCLUSION
dical escalation of cost for acquiring ren a To su urmise the rad e h sourc and post acquirement expenses such as mainteces a esources put wards those re nance upgrades, patches etc. tow e, p e resou urce holder a standpoint thereof. The inevitable sting or alread established computing d dy chang to the exis ges pgrade are also not static for a longer f envir ronment as up hereof the com rces that are mputing resour perio of time. Th od tive of time period will no justify the p ot kept idle irrespect nses that are ost reaso onableness of the overall co and expen ncurring dayny -by-day by an of the reall in ncurred and in ence the conc urce leasing cept of resou sourc owner. He ce
