Biomedical Engineering Reference
In-Depth Information
Fig. 4 Average running time
and collision rate for various
hash algorithms
3.4 Non-cryptographic Model
For some applications, the integrity and authenticity of the data is the only concern
and data con
dentiality is not a problem. This can be achieved by hashing. As
mentioned earlier, whenever a user is registered into the Hadoop database, the user
will be assigned with a secret key and will be converted into hash and given to
Hadoop client node. Whenever the user requested a
le, the Hadoop master node
will generate hash of the requested
le and the respected user
'
s secret key,
appended to the
le and sent to the Hadoop client node.
The Hadoop client node will check the received data, and it will generate its own
hash using the data portion of the received
le and user
'
s secret and will compare it
with received hash, and if both are same, then the
le will be passed to user.
Figure
4
shows the average time needed and collision rate for various algorithms;
even though superfast algorithm is taking less time in execution, its collision rate is
more. Hence, FNV hashing technique is considered and the experimentation is
conducted by using the parameters: running time and collisions.
3.5 Hadoop Cluster Load Monitoring
The size and the load of Hadoop cluster can be very dynamic. Depending upon the
jobs that Hadoop is running, the load of the nodes can be changed. For monitoring
of Hadoop cluster, in this experimentation, Ganglia, an open source software tool, is
used, which offers a complete solution for monitoring, visualizing, and archiving
the system metrics.
4 Evaluation Model
To measure the monitoring system, there are many ways, among which we used
two models to evaluate ECMF, which are accuracy and consumption. We discuss
accuracy model in Sect.
4.1
and consumption model in Sect.
4.2
.
Search WWH ::
Custom Search