Motivating example and problem analysis - Reliability Assurance of Big Data in the Cloud - page 27

of data transfer should not occupy much of the bandwidth over the link as other data
transfers with a higher transfer speed requirement may be happening at the same time.
We refer to this type of data transfer as “lazy” data transfer.
3.2.3.2 Case for data processing within the data center
Many data-intensive Cloud applications need to access large amounts of distributed
data to conduct data-processing tasks. When transferring large amounts of data, these
applications often demand a high data transfer speed. Data transfer could be the major
factor influencing the performance of the application. The time spent on shifting these
data needs to be as short as possible so that the calculation process of the application
is not impacted. In contrast to “lazy” data transfer, we refer to this type of data transfer
as “eager” data transfer. 1
3.2.3.3 Case for across data centers with DCI activities
In spite of data transfer for replica maintenance and data processing in a single data
center, there are also large amounts of data transfer activities with the same purposes
conducted between data centers. Although conducted for the same purposes, these
DCI data transfers are quite different when compared to those conducted within a
data center. Most of such data transfer activities between data centers are dominated
by the lazy style, noninteractive bulk data transfer. These data can range in size from
several terabytes to petabytes [15] . Due to the large data size, both the transfer speed
and energy consumption need to be fully considered. According to one study [74] ,
the shutdown approaches are often used for this type of data transfer to reduce energy
consumption of network devices.
3.2.4 Research issues
The preceding analysis consisted of three parts. First, we analyzed the two major fac-
tors that have led to excessive storage cost for data-intensive applications in current
Cloud storage. Second, we determined on which data storage scheme that our research
should be conducted, so that our motivating example and data-intensive applications
of similar nature could benefit the most. Third, we also analyzed characteristics of the
Cloud network and data transfer activities within the Cloud network. Based on the
analysis, our research issues are further determined with details below.
3.2.4.1 Data reliability model
First of all, we need a model to describe Cloud data reliability and Cloud data reli-
ability-related factors, which is essential for the design of data reliability assurance
1 Here we do not mean that all data transfer for data maintenance is of a “lazy” fashion and all data transfer
for data processing is of an “eager” manner. Some of the data maintenance activities (e.g. many data cre-
ation activities, recovering the data that needs to be processed, etc.) may also need to be conducted in an
eager fashion. Actually, whether the data is transferred in a lazy manner is not determined by the purpose of
the data but the time constraint of the data transfer task and the speed limit of the data transfer.