Project proposal topic:
Markov-based Trace Analysis Algorithm for Wireless Networks
what's the problem? Why is it important?
The internet has become an integral part of our everyday life which provides an enabling environment for business, online shopping, entertainment, bill payment, gaming, etc., as the technology space is dominated by a wireless network, which has been the bedrock for different cloud applications both for business and personal use, which span from video streaming (Netflix, YouTube), to security ( car tracker, phone tracking, intrusion, and fire alarm system), to navigation (GPS, google map), to business ( banking platform setup in cloud, salesforce, and optimization), etc. it is essential to analyze the network behavior through different network algorithm( for this project would be using Markov-based trace analysis) that gives us better insight on the network behavior and performance that would help in optimization and allocation of the network resource to limit delays and errors in a wireless network which leads to scalability, and profitability.
– what has been done?
Research has been done using the Bernoulli model, a memory-less process where each value is generated statistically independent of previous outputs. Thus, this is unlikely to produce accurate models of networks exhibiting burst losses, such as wireless links. Also, some research has been done using the ‘Gilbert Algorithm,’ which uses the Markov chain to model a wireless network, assuming it has a stationary trace which is not a reasonable assumption as it ignores the fact that wireless traces has a time-variant nature.
– what's your approach?
The wireless trace would use the Markov-base trace analysis algorithm (MTA). To achieve this, I would be carrying out the following;
1. partition the trace into similar sub-traces
2. Concatenate the similar sub-traces together to form a stationary trace
3. Associate a state with its new stationary trace
4. Calculate the probability distribution for each state
5. Determine the transition probabilities between states, calculate the mean and standard deviation for the burst length
– expected deliverables and a rough biweekly time schedule
1. Data simulation and collection
2. Statistical analysis of the result
3. Plotting of graphs, comparing results with previous research, and recommendation and conclusion
