Statistics
| Branch: | Revision:

mobicen / documents / old / ExpDefinition.md @ master

History | View | Annotate | Download (2.45 KB)

1 9110387b LoreBz
# Study Pop-Routing in Mobility Scenario
2
3
## Goal
4
5
Understand if, enabling pop-optimazions of Hello-timers on routers that are not fixed  (they are roaming), we still have a gain in terms of route convergence speed. We want to quantify the convergence speed gain. Moreover, as long as this gain depends on the parameters modelling the mobility scenario (e.g.: average nodes velocity...), characterize the gain in function
6
of the mobility parameters.
7
8
## Methodology
9
10
The route convergence speed gain will be defined by comparing simulations of the same mobility scenario
11
using either the optimized or the standard model of a chosen routing protocol.
12
For this purpose, Omnet++ toghether with INET offers a simulation engine for message-passing based protocols that comes
13
with support for mobility and includes also advanced channel models. It also provides some well-known 
14
routing protocols implementations that could be customized to model the pop-optimized version of the protocol that we want to use in the comparative analysis. I guess OSPF (or RIP or BGP) could be the best choice. Inetmanet offers also OLSR, nobody provides Babel.
15
16
A simulation output should be a collection of timestamped RT dumps for each node. For each timestamp the simulator should also
17
log a snapshot of the physical topology of the network (the physical ntw graph). 
18
19
For all smapling instants, we will have the collection of RTs and the topology snapshot: this will let us compute off-line the number of borken paths at that given instant of time. 
20
21
In a simulation of:
22
23
- 5 minutes of duration
24
- with RT sampling every 1 second
25
- with default Hello timer set to 4 second
26
- modelling the space for roaming, the nodes velocity ecc so to have a "neighbour change event" on average every x in {...} sec.
27
    - x could be the "mobility" characteristic parameter
28
29
then we will have a good approximations of the #ofBrokenPahts in function of time for a given x-value of mobility. 
30
31
## Possible Conclusions
32
33
Comparing the #ofBrokenPahts in functions of time for pop-optimized and a standard simulations we can say lot of things, not only average difference between the two.
34
35
After many comparative simulations, for different x-values of mobility, assuming the average difference in the #ofBrokenPahts
36
to be the leading metric, we will be able to determine in which range of the x-values it is worth to apply pop-routing. Moreover
37
we may be even able to understand when mobility is simply too much so that routing never converge...