Synchronous metadata management of large storage systems

Abstract

Today filesystems of big companies are both huge and distributed amongst the world. They contain huge sets of metadata, but are not optimized to analyze them. In contrast, if metadata is stored in a database system and updated synchronously, it could be analyzed and processed in a much easier and straightforward way. Then even adding new attributes, not natively supported by the underlying filesystem, is easily possible. Thus, synchronous metadata storage in a database system can help managing and administrating huge filesystems efficiently but must not slow down the filesystem significantly. The aim of this paper is to describe possible solutions for synchronous metadata storage, inspect how such an integration of filesystem and database system might look like and evaluate the performance.

Referential Integrity in Cloud NoSQL Databases

“…In general, cloud DBMSs require more features than the traditional DBMSs for an efficient data management on the cloud and are found to be less efficient than traditional DBMSs because of the dynamic scalability required to support a changing user-base [2,26,30,36,47,51,52]. Most traditional Relational Database Management Systems (RDBMSs) give data a structure, that adheres to a schema.…”

“…This design was inspired by the experiments done by Hackl et al [30] on a popular NoSQL DBMS named Tokyo Cabinet. Tokyo Cabinet is similar to Cassandra as data is stored in key-value pairs but it does not involve data types or columns and column families as in Cassandra [30,35].…”

“…This design was inspired by the experiments done by Hackl et al [30] on a popular NoSQL DBMS named Tokyo Cabinet. Tokyo Cabinet is similar to Cassandra as data is stored in key-value pairs but it does not involve data types or columns and column families as in Cassandra [30,35]. As a part of their experiments to manage metadata for huge file systems, they adopted an approach to store metadata as a part of the value in a key-value pair.…”