Literature review - Reliability Assurance of Big Data in the Cloud - page 9

Literature review

2

In this chapter, existing literature related to the research are reviewed from three major
aspects. First, to investigate the data reliability pattern in storage devices, we review
literature on hardware reliability theories and existing reliability models of storage de-
vices. Second, to investigate data reliability models and data redundancy maintenance
achieved by using software approaches in the Cloud, literature on data reliability mod-
eling and data reliability assurance approaches in distributed data storage systems
are reviewed. Third, to investigate data transfer for data creation and recovery in the
Cloud, literature on data transfer approaches in distributed systems are reviewed.
The structure of this chapter is organized as follows. In Section 2.1 , we summarize
existing research studies on data reliability assurance in data storage devices in dis-
tributed storage systems. In Section 2.2 , we review research studies on data reliability
assurance with software-based approaches in distributed storage systems, in which we
focus on two major types of data redundancy approaches: data replication and erasure
coding. In Section 2.3 , we review research on data transfer approaches in distributed
systems. Finally, in Section 2.4 , we summarize the works presented in this chapter.

2.1 Data reliability assurance in hardware

In a distributed storage system, there are many factors that could lead to data loss,
such as logical errors that refer to nonphysical failures (e.g., incorrect software up-
dates, programming errors, etc.) and hardware failures (e.g., disk failures), and so
forth [1,16] . However, considering factors that are due to the storage system itself,
hardware failure caused by nonhuman factors is considered to be the major reason
for data loss. No matter how fine the system is designed, the occurrence of hardware
failures is inevitable, where data loss incurs. In that case, the reliability of data is de-
termined by the storage device on which the data are stored. In this section, we review
existing research studies on data reliability theories for storage devices.

2.1.1 Disk
It is estimated that more than 90% of all new information produced in the world is
stored on magnetic media, mostly on hard disk drives [17] . In current Clouds, disks
are still the most commonly used storage device for storing the massive amount of
Cloud data. Investigations toward the reliability pattern of disks have been conducted
for decades in both academia and industry [2,18-20] . With the development of dis-
tributed systems, such as clusters [21] , peer-to-peer (P2P) systems [22,23] , grids [7] ,
and Clouds [24-26] , the reliability issues of disks become more important compared
to systems with centralized storage due to the big disk amount and more disk failures.