Biomedical Engineering Reference
In-Depth Information
4.1 Accuracy
The accuracy of the monitoring information gives the error between the collected
and the actual data.
Assume there are N metrics available at system that we want to monitor.
X
1
, X
2
,
…
X
N
are the maximum measure of the N metrics.
V
1
, V
2
,
V
N
are the collected measure of the N metrics.
A
1
, A
2
,
…
A
N
are the actual measure of the N metrics.
Then, the accuracy of the monitoring information is
…
N
X
N
1
½
V
i
A
i
A
¼
1
X
i
i¼1
The metrics are memory, CPU and disk usage, and workload on the network.
4.2 Consumption
The task of monitoring middleware will consume resources. The overall usage can
be given as the total cost of CPU, memory, disk usages, and workload on network,
represented as (u
1
), (u
2
), (u
3
), and (u
4
), respectively.
X
4
U
i
T
cons
¼
P
i
i¼4
where P
i
is the cost of the Amazon EC2 pricing that has been referenced from.
5 Experimental Results
5.1 Environment Setup
To verify performance of proposed system, a Hadoop cluster of one name node and
three data nodes is selected. They are con
gured with 2.6 GHz processor and 4 GB/
500 GB and used high-speed switches to connect all these. The OS used is Linux
4.1, and version of Hadoop used is 0.20.1, and for GNU, Java 1.6.0 is installed in
all nodes. For virtualization and establishing infrastructure, we used Xen hypervisor
and eucalyptus. And for performance monitoring, Ganglia-3.6.0 is used.
Search WWH ::
Custom Search