![]() | Only 14 pages are availabe for public view |
Abstract Today big data represent a major challenge for the performance of cloud computing storage systems. Some Distributed File Systems (DFS) are widely used to store big data, such as Hadoop Distributed File System (HDFS), Google File System (GFS) and others. These DFS replicate and store data as multiple copies to achieve high availability and reliability, but they increase storage and resources consumption. Security is considered one of the most critical aspects in a cloud computing environment. Users of these cloud services usually store sensitive and important information, therefore the trustworthiness of the providers is so important. Single cloud providers, for example, are predicted to become less popular due to the inherent risks of service availability, data lock-in and the possibility of malicious attacks from insiders and outsiders. Most users are migrating toward multi-cloud service providers. This research addresses these issues by presenting three approaches for multi-cloud storage management; Redundant Independent Files (RI F), Secure Distributed Redundant Independent Files (SDRIF) and High Availability Redundant Independent Files (HARIF). It also suggests the use of two encryption algorithms to solve the problem of data confidentiality; simple encryption mode and complex encryption mode. These two algorithms are used with the SDRIF approach. The RIF approach is a technique for reducing storage and resources in big data replication. The RIF is a service layer built above the cloud providers (CP) without changing the roperties of HDFS. It splits data into three parts and uses the XOR operation to generate a fourth part as parity. These four parts are stored in HDFS files as independent files. The generated parity file not only guarantees the integrity, =vailability aprj reliability of data but also reduces storage space, resources consumption a.. uperai.. costs. When RIF is combined with the CP, the produced model CPRIF improves the reading and writing performance compared to other models. The SDRIF addresses some issues found with the RIF approach and it mainly if troduced data confidentiality that the RIF approach lacks to offer. It works similar to RIF, but the enerated parity is not stored in one separate file. The generated parity blocks are distributed among all data parts. The CPSDRIF is the model produced when combining the SDRIF with CP. To offer data confidentiality, two encryption algorithms are used to provide secure DFS by injecting an encryption/decryption layer between SDRlF and CP. This layer performs data encryptionldecryption with two modes; simple encryption mode using an XOR operation and complex encryption mode using the AES-CTR algorithm. The HARlF approach is suggested to overcome availability and reliability limitations, time overheads of recovering data and to increase data integrity of the RIF. The CPHARIF is the name of the model when combining the HARIF approach with the CP and it offers these improvements along with all the benefits that the CPRIF model offers. The key idea of the proposed CPHARlF model is to increase the replication of data before it is stored in HDFS files to achieve better availability, integrity and reliability. It splits data into three parts and uses the XORing operation to generate three parity parts that provide data integrity and availability. The suggested models were implemented on servers running multiple virtual nodes and with the TeraGen benchmark tool, different performance aspects were tested. The results show that the suggested CPRIF model decreased the storage space by 33% compared to other models and improved the data reading and writing by about 34%. The results show that the CPSDRlF model has improved the data scurity compared to CPRIF. Also, CPSDRlF consumed the storage space by 67% compared to other models and improved the data reading and writing by about 34%. The results also show that the CPHARIF model has improved the availability, integrity, reliability and data recovey time compared to CPRIF. The CPHARIF reduces stored data size and resources consumption, improves writing operation upto 7% compared to the other models. Generally, big data storage is saved and 10 Operations performance is improved. Keywords: Big Data, Cloud Storage, Cloud Computing, Google File System (GFS), HDFS, Availability, Integrity, Confidentiality, AES-CTR. |