Storage Systems for Big Data
Internet age comes with the vast amount of data that requires efficient storage and processing capabilities. To alleviate this issue, we discuss data storage systems which are tailored to store and process big data effectively. While general-purpose RDBMSs are still viable option in handling and analyzing structural data, they suffer from a variety of problems including performance and scalability issues when it comes to big data. To increase the performance of DBMS for big data storage needs, partitioning the data across several sites and paying big license fees for enterprise SQL DBMS might be the two possible options (Stonebraker, 2010), however they are not even without disadvantages such as inflexible data …show more content…
Then Bigtable returns the corresponding data where it behaves similar to distributed hash table. Bigtable maintains the data in lexicographic order by row string and the data is partitioned according to the row range, called a tablet. By incorporating the timestamp value, Bigtable gains the ability to store multiple versions of the same data. For faster retrieval of the tablets, Bigtable takes advantage of a three-level hierarchy analogous to …show more content…
In recent years, researchers have begun to seek efficient ways to outperform legacy database systems. As the RAM capacities increase, the technique of storing partitions of data on the RAM of shared-nothing machines is more applicable than ever. NewSQL databases are designed by taking advantage of some modern techniques such as data sharding, data replication and distributed memory database and offer scalable and high performance solution to disk-based legacy database systems. NewSQL databases provide an object oriented database language that is considered easier to learn than the standard SQL language (Kumar et al., 2014).
H-Store (Kallman et al., 2008) divides database into partitions where each partition is replicated and resides in main memory. The H-Store system relies on distributed machines that share no data to improve the overall performance of database
Internet age comes with the vast amount of data that requires efficient storage and processing capabilities. To alleviate this issue, we discuss data storage systems which are tailored to store and process big data effectively. While general-purpose RDBMSs are still viable option in handling and analyzing structural data, they suffer from a variety of problems including performance and scalability issues when it comes to big data. To increase the performance of DBMS for big data storage needs, partitioning the data across several sites and paying big license fees for enterprise SQL DBMS might be the two possible options (Stonebraker, 2010), however they are not even without disadvantages such as inflexible data …show more content…
Then Bigtable returns the corresponding data where it behaves similar to distributed hash table. Bigtable maintains the data in lexicographic order by row string and the data is partitioned according to the row range, called a tablet. By incorporating the timestamp value, Bigtable gains the ability to store multiple versions of the same data. For faster retrieval of the tablets, Bigtable takes advantage of a three-level hierarchy analogous to …show more content…
In recent years, researchers have begun to seek efficient ways to outperform legacy database systems. As the RAM capacities increase, the technique of storing partitions of data on the RAM of shared-nothing machines is more applicable than ever. NewSQL databases are designed by taking advantage of some modern techniques such as data sharding, data replication and distributed memory database and offer scalable and high performance solution to disk-based legacy database systems. NewSQL databases provide an object oriented database language that is considered easier to learn than the standard SQL language (Kumar et al., 2014).
H-Store (Kallman et al., 2008) divides database into partitions where each partition is replicated and resides in main memory. The H-Store system relies on distributed machines that share no data to improve the overall performance of database