CloudDB AutoAdmin - Large Scale and Big Data: Processing and Management - page 379

11.6.2.2 Replication Delay
Figure 11.10 illustrates the effect of the adaptive replication controller on the perfor-
mance of the replication delay for the cloud-hosted database replicas. Figure 11.9a
and b shows the replication delay of the two baseline cases for our comparison. They
represent the experiments of running with a fixed number of replicas (3 and 6, respec-
tively) from the starting times of the experiments to their end times. Figure 11.9a shows
that the replication delay tends to follow different patterns for the different replicas.
The two trends of us - west -1 and eu - west -1 surge significantly at 260 and 280 users,
respectively. At the same time, the trend of us - east -1 tends to be stable throughout the
entire running time of the experiment. The main reason behind this is the performance
variation between the hosting EC2 instances for the database replicas.* Due to the
performance differences between the physical CPUs specifications, us - east -1 is able to
handle the amount of operations that saturate us - west -1 and eu - west -1. Moreover, with
an identical CPU for us - west -1 and eu - west -1, the former seems to surge at an earlier
point than the latter. This is basically because of the difference in the geographical
location of the two instances. As illustrated in Figure 11.8, the MySQL Proxy location
is closer to us - west -1 than eu - west -1. Therefore, the forwarded database operations by
the MySQL Proxy take less time to arrived at us-west- 1 than eu-west- 1, which leads to
more congestion on the us - west -1 side. Similarly, in Figure 11.9b, the replication delay
tends to surge in both us - west -1 and us - west -2 for the same reason of the difference in
the geographic location of the underlying database replica.
Figures 11.9c and 11.10d-f show the results of the replication delay for our experi-
ments using different values for the monitor interval ( intvl mon ) and the tolerance of
replication delay ( delay tolerance ) parameters. For example, Figure 11.9c shows that
the us - west -2, us - east -2, and eu - west -2 replicas are added in sequence at the 255th,
407th, and 1843th seconds, where the drop lines are emphasized. The addition of
the three replicas are caused by the SLA-violation of the us - west -1 replicas at differ-
ent periods. In particular, there are four SLA-violation periods for us - west -1 where
the period must exceed the monitor interval, and all calculated replication delays
in the period must exceed the SLA of replication delay. These four periods are:
(1) 67:415 (total of 349 seconds). (2) 670:841 (total of 172 seconds). (3) 1373:1579
(total of 207 seconds). (4) 1615:3000 (total of 1386 seconds). The addition of new
replicas is only triggered on the first and the fourth periods based on the time point
analysis. The second and the third periods do not trigger the addition of new replica
as the number of detected SLA violations does not exceed the defined threshold ( T ).
Figures 11.9c and 11.10a-c show the effect of varying the monitor interval ( intvl mon )
on the replication delay of the different replicas. The results show that us - west -2 is always
the first location that adds a new replica because it is the closest location to us - west -1,
which hosts the replica that first violates its defined SLA data freshness. The results also
show that as the monitor interval increases, the triggering points for adding new replicas
are usually delayed. On the contrary, the results of Figures 11.9c and 11.10d-f show that
increasing the value of the tolerance of the replication delay parameter ( delay tolerance ) does
not necessarily cause a delay in the triggering point for adding new replicas.
* Both us-west- 1 and eu-west- 1 are powered by Intel ® Xeon ® CPU E5507 at 2.27GHz, whereas us-east- 1
is deployed with a better CPU, Intel ® Xeon ® CPU E5645 @ 2.40GHz.