Database Reference
In-Depth Information
called a scan cycle , in which all of the metadata in the data table are sequentially
scanned once. To proactively check all the managed data files in a certain frequency,
the time for the scan cycle of each PRCR node is set to a fixed value. By doing so, the
scanning frequency of metadata can be determined so that proactive replica checking
tasks can be conducted well in time. However, due to the limited performance of the
compute instance that the PRCR node is running on, the time constraint of the scan
cycle means that the maximum capacity of the PRCR node is also limited.
Replica management module : It is responsible for scanning the metadata in the data
table and cooperating with the Cloud compute instances to process the proactive replica
checking tasks. In each scan cycle, the replica management module scans the metadata
in the data table and determines whether the data file needs to be checked. If a data file
needs to be checked, the replica management module obtains its metadata from the data
table and sends it to a Cloud compute instance for proactive replica checking. After the
proactive replica checking task is finished, the replica management module conducts
further actions according to the returned result. In particular, if any replica is lost, the
replica management module initializes the recovery process for creating a new replica.
For the recovery of data files with different usages, contents, and formats, different
data recovery strategies can be applied to achieve certain goals. As will be presented in
Chapter 7 , our data transfer approach for data creation and data recovery can be applied
to reduce the energy consumption during the data recovery process.
6.3
Working process of PRCR
By tightly integrating all the functions of PRCR components, a series of activities for
maintaining Cloud data files with data reliability assurance is conducted, in which the
complex working process of PRCR is formed. In Figure 6.2 , we illustrate this process
by following the life cycle of a data file managed by PRCR in the Cloud.
1. The process starts at the time that the original replica of the data file is created in the Cloud.
According to the minimum replication approach, the user interface determines the minimum
number of replicas, that is, to store the data file with one replica or two replicas. Specifically,
Figure 6.2 Working process of PRCR
 
Search WWH ::




Custom Search