Image Processing Reference
In-Depth Information
3
1
2
5
A
6
4
7
8
9
D
10
B
11
C
Fig. 12.2 Network topology example. End users A-D form a social network
apply classical routing algorithms, such as spanning tree creation/refinement, in
order to identify the optimum routes of service delivery (ALM service nodes, cache
content placement nodes) that would best service the identified user community.
The operation rationale behind this concept at service level is to have background
processes implementing the above workflow, identifying social communities and
pre-calculating for them routing trees, cache them and deploy them when temporal
or session-driven trigger conditions are met (e.g. login of N1 out of N members of
the community).
The main advantage in this approach is that it breaks down the problem of
identifying optimum content placement strategies and multicast trees, to multiple
smaller problems involving, small groups of nodes, of which the statistical proba-
bility of its usefulness can be quantified based on historic data. Furthermore, data
processing and algorithm execution is asynchronous without real-time requirement.
It is a task executed in the background and could be delegated to distributed
computing frameworks, such as Map Reduce [ 13 - 15 ] containers. It is worth
mentioning that open platforms such as Apache Hadoop (MapReduce container
[ 13 ]) and Apache Mahout (Data Mining Big Data Engine [ 14 ]) have evolved
significantly so that this level of processing complexity can be feasible at a high
programming level and low investment cost.
 
Search WWH ::




Custom Search