Published on

  • Be the first to comment

  • Be the first to like this

No Downloads
Total views
On SlideShare
From Embeds
Number of Embeds
Embeds 0
No embeds

No notes for slide


  1. 1. B.AmarNadh Reddy, P.Raja Sekhar Reddy / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 2, Issue 5, September- October 2012, pp.1130-1134 Effective Data Distribution Techniques for Multi-Cloud Storage in Cloud Computing B.AmarNadh Reddy*1,P.Raja Sekhar Reddy2 *1CSE, Anurag Group of Institutions, Hyderabad, A.P, India 2 CSE, Anurag Group of Institutions, Hyderabad, A.P, IndiaAbstract Cloud data storage redefines the issues issues with large amounts of data storage. Intargeted on customer’s out-sourced data (data addition to these benefits, customers can easilythat is not stored/retrieved from the costumers access their data from any geographical regionown servers). In this work we observed that, where the Cloud Service Provider’s network orfrom a customer’s point of view, relying upon a Internet can be accessed [1]. An example of thesolo SP for his outsourced data is not very cloud computing is shown in Fig. 1. Since cloudpromising. In addition, providing better privacy service providers (SP ) are separate market entities,as well as ensure data availability and reliability data integrity and privacy are the most critical issuescan be achieved by dividing the user’s data block that need to be addressed in cloud computing. Eveninto data pieces and distributing them among the though the cloud service providers have standardavailable SP s in such a way that no less than a regulations and powerful infrastructure to ensurethreshold number of SP s can take part in customer’s data privacy and provide a bettersuccessful retrieval of the whole data block. In availability, the reports of privacy breach andthis paper, we propose a traditional technique service outage have been apparent in last few yearsfollowed to distribute data to multi-cloud storagemodel in cloud computing which holds aneconomical distribution of data among theavailable SPs in the market, to provide customerswith data availability as well as reliability . Datafragmentation plays an important role in datadistributionKeywords Cloud computing, storage, Cloud serviceprovider, customer .Fragmentation I. INTRODUCTION The end of this decade is marked by a paradigmshift of the industrial information technology Fig. 1. Cloud computing architecture exampletowards a subscription based or pay-per-use servicebusiness model known as cloud computing. This In this work we observed that, from a customer’sparadigm provides users with a long list of point of view, relying upon a solo SP for hisadvantages, such as provision computing outsourced data is not very promising. In addition,capabilities; broad, heterogeneous network access; providing reliability as well as ensure dataresource pooling and rapid elas-ticity with measured availability, can be achieved by dividing the user’sservices .Huge amounts of data being retrieved from data block into data pieces and distributing themgeographically distributed data sources, and non- among the available sp’s.localized data-handling requirements, creates such achange in technological as well as business model. To address these issues in this paper, we proposedOne of the prominent services offered in cloud the techniques for distribution of data among thecomputing is the cloud data storage, in which, available SP s in the market, to provide customerssubscribers do not have to store their own data on with data availability as well as reliability [1]. In ourtheir servers, where instead their data will be stored model, the customer divides his data among severalon the cloud service provider’s servers. In cloud SP s available with respect to data access quality ofcomputing, subscribers have to pay the provides for service offered by the SP s at the location of datathis storage service. This service does not only retrieval. This not only rules out the possibility of aprovides flexibility and scalability data storage, it SP misusing the customers’ data, breaching thealso provides customers with the benefit of paying privacy of data, but can easily ensure the dataonly for the amount of data they needs to store for a availability with a better quality of service.particular period of time, without any concerns of Customers’ stored data at cloud service providers isefficient storage mechanisms and maintainability vulnerable 1130 | P a g e
  2. 2. B.AmarNadh Reddy, P.Raja Sekhar Reddy / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 2, Issue 5, September- October 2012, various threats. Previous studies in [9], [11] cloud service providers are available for customerdiscussed in detail that a cloud service provider can (C1), who want to store his own data securely. Inbe a victim to Denial here he will divide his data into two parts (D1 andof service attacks or its variants. D2) and distribute these parts on the two available CSPs (CSP1 and CSP2) respectively. The two cloudwe consider two types of threat models. First is the service providers might collude with each other, andsingle point of failure [5], [6], which will affect the Exchange the parts of data that the customer hasdata availability, that could occur if a server at the stored on their server and reconstruct the whole datacloud service provider failed or crashed, which without being detectedmake it hard for the costumer to retrieve his stored by the userdata from the server. Availability of data is also animportant issue which could be affected, if the cloud Our proposed approach will provide the cloudservice provider (SP) runs out of business. Such computing users a decision model, that provides aworries are no more hypothetical issues, therefore, a better reliability and availability by distributing thecloud service customer can not entirely rely upon a data over multiple cloud service providers in such asolo cloud service provider to way that, none of the SP can successfully retrieveensure the storage of his vital data. and use it. II. DATA DISTRIBUTION Data preservation and data integrity are two of the most critical security issues related to user data[2],[11]. In conventional paradigm, the organizations had the physical possession of their data, and thus have an ease of implementing better data availability policies. But in case of cloud computing, the data is stored on an autonomous business party, that provides data storage as a subscription service. The users have to trust the cloud service provider (SP ) with security of theirTo illustrate this threat we use an example in Fig. 2. data. In the author discussed the criticality of theLet us assume that three customers (C1, C2 and C3) privacy issues in cloud computing, and pointed outstored their data on three different service providers that obtaining an information from a third party is(CSP1, CSP2 and CSP3) respectively. Each much more easier than from the creator himself.customer can retrieve his own data from the cloud One more bigger concern that arises in suchservice provider who it has a contract with. If a schemes of cloud storage services, is that, there is nofailure full-proof way to be certain that the service provideroccur at CSP1, due to internal problem with the doe not retains the user data, even after the user optsserver or some issues with the cloud service out of the subscription. With enormous amount ofprovider, all C1’s data which was stored on CSP1’s time, such data can be decrypted and meaningfulservers will be lost and cannot be retrieved. One information can be retrieved and user privacy cansolution for this threat is that, the user will seek to easily be breached. In order to stop the SP tostore his data at multiple service providers to ensure observe the data ,data can be fragmented andbetter availability of his data. distributed to several SP’s .Our second threat discussed in this paper is thecolluding service providers, in which the cloud Why Fragment?service providers might collude together toreconstruct and access the user stored • Usagedata.In [9] the authors provide the idea for distributing • Applications work with views rather thanthe data among two storage clouds such that, an entire relations.adversary cannot retrieve the contents of the data • Efficiencywithout having access toboth the storage clouds. • Data is stored close to where it is mostRelaying entirely upon a couple of service providers frequently used.for the storage and retrieval of data might not be • Data that is not needed by local applicationssecured against colluding service providers. Such an is not storedattack scenario is entirely passive, because the clouduser cannot detect that his information has been • Parallelismcollectively retrieved from the service providerswithout his consent. We illustrate the colluding • With fragments as unit of distribution,service providers’ threat .Let us assume that two transaction can be divided into several 1131 | P a g e
  3. 3. B.AmarNadh Reddy, P.Raja Sekhar Reddy / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 2, Issue 5, September- October 2012, pp. sub-queries that operate on fragments. [11] • Πa1,…an(R) • Security • Determined by establishing affinity of one attribute to another. Data not required by local applications is notstored and so not available to unauthorized users • Example:Types of Fragmentation • Relation: • Four types of fragmentation: Bars(name,address,licence,employees,owner ) • Horizontal, [11] • Fragments: • Vertical, » Πname,address,licence (Bars) • Mixed, » Πname,address,employees,owner(Bars) • Derived. Mixed Fragmentation • Other possibility is no fragmentation: • We can also mix horizontal and vertical If relation is small and not updated frequently, fragmentation.may be better not to fragment relation • We obtain a fragment that consist of an horizontal fragment that is Horizontal Fragmentation vertically fragmented, or a vertical fragment that is horizontally • Each fragment consists of a subset of the fragmented. tuples of a relation R. [11] • Defined using Selection operation of relational • Defined using Selection and algebra: Projection operations of relational algebra. • σp(R) • Example: σ (Π p a1,…an) Πa1,…an(σp) • Relation: Sells(pub, address,price,type) Derived Horizontal Fragmentation • Fragments: • A horizontal fragment that is based on horizontal fragmentation of a parent » SellsBitter= σtype = “bitter”(Sells) relation.[11] » SellsLager= σtype = “lager”(Sells) • Ensures that fragments that are frequently This strategy is determined by looking at joined together are at same site. predicates used by transactions. • Involves finding set of minimal (complete • Defined using Semijoin operation of and relevant) predicates. relational algebra: • Set of predicates is complete, if and only if, any two tuples in same fragment are Ri = R >F Si, 1≤i≤w referenced with same probability by any • If relation contains more than one application. foreign key, need to select one as parent. • Predicate is relevant if there is at least one application that accesses fragments • Choice can be based on fragmentation differently. used most frequently or fragmentation with better join characteristics. Vertical Fragmentation How can we define fragments correctly In defining fragments we have to be very careful. • Each fragment consists of a subset of attributes of a relation R. [11] Three correctness rules: Completeness: • Defined using projection operation of relational algebra: If relation R is decomposed into fragments r1, r2, …rn, each data item that 1132 | P a g e
  4. 4. B.AmarNadh Reddy, P.Raja Sekhar Reddy / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 2, Issue 5, September- October 2012, pp. can be found in R must appear in at least one fragment. This ensures no loss of data Reconstruction: The Bars relation can be during fragmentation[11] reconstructed from the fragments Disjointness: The two fragments are disjoint,Reconstruction: except for the primary key, name, which iswe must be able to reconstruct the entire R necessary for reconstructionfrom fragments.For horizontal fragmentationis union operation.[11] Our model distributes the data pieces among more than one service providers, in such a way that no R = r1 ∪ r2 ∪ … ∪ rn, one of the SP s can retrieve any meaningful For vertical fragmentation is natural information from the pieces of data stored on itsjoin operation. R = r1 >< r2 >< … >< servers, without getting some more pieces of datarn, from other service providers. Therefore, the conventional single service provider basedTo ensure reconstruction we have to include primary techniques does not seem too much promising.key attributes in all fragments Distributing the data over multiple clouds orDisjointness networks in such a way that if an adversary is able to intrude in one network, still he cannot retrieveIf data item x appears in fragment ri, then it should any meaningful data, because its complementarynot appear in any other fragment. pieces are stored in the other network. Our approach is similar to this approach, because both aim toException: vertical fragmentation, where remove the centralized distribution of cloud data.primary key attributes must be repeated Although, in their approach, if the adversary causesto allow reconstruction. For horizontal a service outage even in one of the data networks,fragmentation, data item is a tuple For the user data cannot be retrieved at all. This is whyvertical fragmentation, data item is an in our model; we propose to use a redundantattribute distribution scheme in which at least a thresholdCorrectness of Horizontal Fragmentation number of pieces of the data are required out of the entire distribution range, for successful retrieval.Relation:Sells(pub,address,price,type)type={Bitter, Lager} Fragments: Meaningful information from the data pieces • SellsBitter= σtype = “bitter”(Sells) allocated at their servers. Also, in addition, we • SellsLager= σtype = “lager”(Sells) provide the user with better assurance of availability of data, by maintaining redundancy in dataCorrectness rules distribution. In this case, if a service provider suffers service outage [1] [12] or goes bankrupt, the user Completeness: Each tuple in the relation still can access his data by retrieving it from otherappears either in SellsBitter, or in SellsLager service providers. From the business point of view, since cloud data Reconstruction: The Sells relation can be storage is a subscription service, the higher the datareconstructed from the fragments redundancy, the higher will be the cost to be paid by Sells = SellsBitter X SellsLager the user. III. MODELSDisjointness: The two fragments are disjoint, we will describe system model and thethere can be no beer that is both “Lager” and distribution model. The terms cloud storage and“Bitter” Correctness of Vertical Fragmentation cloud data storage are interchangeable, also theRelation: terms user and customer are interchangeable.Bars(name,address,licence,employees,owner) System OverviewFragments: We consider the storage services for cloud data r =Π 1 name,address,licence (Bars) storage be-tween two entities, cloud users (U) and r =Π cloud service providers 2 name,address,employees,owner(Bars)Correctness rules (SP ). The cloud storage service is generally priced on two factors, how much data is to be stored on the Completeness: Each attribute in the Bars cloud servers and for how long the data is to berelation appears either in r1 or in r2 stored. In our model, we assume that all the data is to be stored for same period of time. 1133 | P a g e
  5. 5. B.AmarNadh Reddy, P.Raja Sekhar Reddy / International Journal of Engineering Research and Applications (IJERA) ISSN: 2248-9622 Vol. 2, Issue 5, September- October 2012, pp. We consider p number of cloud service providers Privacy Forum, online at(SP ), each available cloud service provider is with a QoS factor, along with its cost of PF Cloud Privacy Report.pdf,Feb 2009.providing storage service per unit of stored data (C). [5] N. Gruschka, M. Jensen, “Attack surfaces:Every SP has a different level of quality of service A taxonomy for attacks on cloud services”,(QoS) offered as well as a different cost associated Cloud Computing (CLOUD), 2010 IEEEwith it. Hence, the cloud user can store his data on 3rd International Conference on, 5-10 Julymore than one SP s according to the required level 2010.of security and their affordable budgets. [6] W. Itani, A. Kayssi, A. Chehab, “Privacy as a Service: Privacy-Aware Data StorageIV. Distribution Model and Processing in Cloud Computing Customers’ stored data at cloud service Architectures,” Eighth IEEE Internationalproviders is vulner-able to various threats. Previous Conference on Dependable, Autonomic andstudies in a cloud service provider can be a victim to Secure Computing, Dec 2009Denial of service attacks or its variants[3],[4]. [7] M. Jensen, J. Schwenk, N. Gruschka, L.L.The idea for distributing the data among two storage Iacono, “On Technical Security Issues inclouds such that, an adversary cannot retrieve the Cloud Computing”, IEEE Internationalcontents of the data without having access to both Conference on Cloud Computing, (CLOUDthe storage clouds. Relaying entirely upon a couple II 2009), Banglore, India, September 2009,of service providers for the storage Such an attack 109-116scenario is entirely passive, because the cloud user [8] J. Kincaid, “MediaMax/TheLinkup Closescannot detect that his information has been Its Doors”, Online atcollectively retrieved from the service providers his consent.. Let us assume that two cloud iamaxthelinkup-closes-itsdorrs/, July 2008.service providers are available for customer who [9] P. F. Oliveira, L. Lima, T. T. V. Vinhoza, J.want to store his own data securely. seeks a Barros, M. M´edard,“Trusted storage overdistribution of customer’s data pieces among the untrusted networks”, IEEE GLOBECOMavailable SP s in such a way that, at least q number 2010, Miami, FL. USA.of SP s must take part in data retrieval, while [10] S. H. Shin, K. Kobara, “Towards secureminimizing the total cost of storing the data on SP s cloud storage”, Demo forCloudCom2010,as well as maximizing the quality of service and Dec 2010.availability of data provided by the SP s. [11] Stefano ceri Giuseppe pelagati “Distributed Databases-Principles and V. CONCLUSION systems”Tata MC Graw Hill 2008 In this paper, we proposed a different data [12] J. Du, W. Wei, X. Gu, T. Yu, “RunTest:fragmentation schemes for multi cloud storage in assuring integrity of dataflow processing incloud computing, which seeks to provide each cloud computing infrastructures”, Incustomer with reliability, availability and better Proceedings of the 5th ACM Symposium oncloud data storage decisions. Information, Computer and Communications Security (ASIACCS ’10),ACKNOWLEDGEMENT ACM, New York, NY, USA, 293-304 We are very much thankful to Prof. G.Vishnu Murthy, Head of the CSE Dept who hadgiven valuable suggestions in carrying out thispaper.REFERENCES [1], “Amazon s3 availablity event: July 20, 2008”, Online at 20080720.html, 2008. [2] P. S. Browne, “Data privacy and integrity: an overview”, In Proceeding of SIGFIDET ’71 Proceedings of the ACM SIGFIDET (now SIGMOD), 1971. [3] A. Cavoukian, “Privacy in clouds”, Identity in the Information Society, Dec 2008. [4] R. Gellman, “Privacy in the clouds: Risks to privacy and confidentiality from cloud computing”, Prepared for the World 1134 | P a g e