Database Reference
In-Depth Information
deployed in three regions, namely:
us-west
,
us-east
and
eu-west
. All slaves run in
small instances for the same reason of provisioning the master instance.
Two sets of experiments are implemented in order to evaluate the effectiveness
of the SLA management framework in terms of its effectiveness on maximizing
the end-to-end system throughput and minimizing the replication delay for the
underlying virtualized database servers. In the first set of experiments, the value
of the tolerance component,
delay
tolerance
, of the SLA replication delay is fixed at
1,000 ms, and the monitor interval,
intvl
mon
, is varied among the following set of
values, 60, 120, 240, and 480 s. In the second set of experiments, in contrast to the
first test, the monitor interval, int
v
l
mon
, is fixed at 120 s, and the SLA of replication
delay is adjusted by varying the tolerance component of the replication delay,
delay
tolerance
, among the values of 500, 1,000, 2,000, and 4,000 ms. In the experiment
environment, the round-trip component for the virtualized database replica servers
is determined with
ping
command running every second for a 10 min period. The
average round-trip time of three geographical regions is 30, 130, and 200 ms from
the master to slaves in
us-west
,
us-east
, and
eu-west
respectively.
Every experiment runs for a period of 3,000 s with a starting workload of
220 concurrent users and database requests with read/write ratio at 80/20. The
workload gradually increases in steps of 20 concurrent users every 600 s so that
each experiment ends with a workload of 300 concurrent users. Each experiment
deploys 6 virtualized database replica servers in 3 regions where each region hosts
two replica servers: the first replica server is an active replica which is used from the
start of the experiment for serving the database requests of the application while the
second one is a hot backup which is not used for serving the application requests at
the beginning of the experiment but can be added by the action module, as necessary,
when triggered by the control module. Finally, in addition to the two sets of
experiments, two experiments without the adaptive SLA management framework
are conducted as baselines for measuring the end-to-end throughputs and replication
delays of 3 and 6 slaves, representing the minimum and the maximum number of
running replica servers, respectively. For all experiments, the value of the heartbeat
interval,
intvl
heart
, is set to 1 s and the value of the threshold, T , for the maximum
possible continuous SLA violations for any replica server is calculated using the
following formula T
int
v
l
mon
D
int
v
l
heart
.
7.4
Evaluation of SLA Management Framework
End-to-End Throughput
Table
7.1
presents the end-to-end throughput results for the set of experiment with
different configuration parameters. The baseline experiments represent both the
minimum and the maximum end-to-end throughput results with 22:33 and 38:96
operations per second respectively. The end-to-end throughput delivered by the
