Database Reference
In-Depth Information
servers in cloud environment. In particular, this chapter focuses on addressing the
following questions with regards to the
master-slave
database replication strategy
on Amazon EC2:
How well does the master-slave replication strategy scale with an increasing
workload and an increasing number of virtualized database replica servers in
cloud? In principle, we try to understand what factors act as limits on achievable
scale.
What is the average replication delay or window of data staleness that could exist
with an increasing number of virtualized database replica servers and different
configurations to the geographical locations of the slave databases?
The remainder of this chapter is structured as follows. In Sect.
6.1
, a few design
decisions that are related to the benchmark application are explained, including
customizing Cloudstone implementing fine-grained time/date function in MySQL,
and applying clock synchronization in cloud. Meanwhile, Sect.
6.2
details the
implementation of the experimental framework and the experimental environment.
While the results of our experiments are presented in Sect.
6.3
. Finally, the
conclusion of the experiments are discussed Sect.
6.4
.
6.1
Design of Benchmark Application
Figure
6.1
shows the overall architecture of relational database as a service
benchmark application. In general, it is a three-layer implementation. The first
layer is a customized Cloudstone benchmark [
1
] which controls the read/write
ratio and the workload. The second layer includes a master database that receives
write operations from the benchmark and is responsible for propagating
writesets
to
slaves. The third layer is a group of slaves which are responsible for processing read
operations and updating writesets.
The design of the benchmark tool is relational-database-focused and replication-
precision-driven [
242
]. Therefore, there are several issues need to be addressed
during the design of the benchmark application. Such as enforcing Cloudstone
to benchmark database tier only, enabling the ability of benchmarking replication
delay, tweaking time/date function in MySQL for precious resolution to calculate
a replication delay, and enforcing clock synchronizations. All the detailed design
decisions are discussed as following.
Customized Cloudstone
The Cloudstone benchmark has been designed as a performance measurement tool
for Web 2.0 applications. The benchmark mimics a Web 2.0 social events calendar
that allows users to perform individual operations such as browsing, searching,
