Video Traffic SLA Prediction Using Markovian Arrival Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for predicting Service Level Agreement (SLA) performance for video service-centric data traffic are inadequate due to the non-compliance of video traffic with classic Poisson models, leading to inaccurate predictions and suboptimal network planning and resource allocation.
Innovation Solution
An SLA performance prediction method using a Markovian arrival process (MAP) model to determine time and packet length feature parameters, enabling accurate prediction of latency and cache queue length by calculating cumulative probability distributions based on these parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a classic Poisson model is used to model video service-centric data traffic, then the modeling process is simple, but the prediction accuracy deteriorates because video traffic does not comply with Poisson model assumptions
Solution Approach 1:
The patent changes the mathematical model parameters from a simple Poisson process to a Markovian Arrival Process (MAP) model, which introduces additional state variables and transition parameters. This parameter change allows the model to capture the bursty and correlated nature of video traffic, thereby improving prediction accuracy while maintaining manageable complexity through systematic parameter estimation.
Solution Approach 2:
The patent introduces an intermediary Markovian state space between the observed traffic data and the SLA performance predictions. The MAP model acts as a mediator that processes raw traffic measurements (inter-arrival times and packet lengths) into meaningful traffic characteristics, which then feed into the SLA prediction calculations for latency and queue length.
2Measurement precision
If a Markovian arrival process (MAP) model is used to model video traffic, then the prediction accuracy improves, but the device complexity increases due to additional modeling requirements
Solution Approach 1:
The patent segments the complex MAP modeling process into distinct, manageable modules: (1) collecting raw traffic data (timestamps and packet lengths), (2) calculating inter-arrival times and packet length statistics, (3) estimating MAP model parameters (transition rates and probabilities), and (4) computing SLA performance predictions. This segmentation reduces the perceived complexity by breaking down the overall process into sequential, well-defined steps.
Solution Approach 2:
The patent uses simulated data generated from the fitted MAP model to represent actual video traffic characteristics. By creating synthetic traffic streams that replicate the statistical properties of real video traffic, the system can perform SLA predictions without needing to process complex real-time traffic data directly, thereby reducing computational complexity while maintaining prediction accuracy.
3Ease of manufacture
If conventional voice service traffic modeling methods are used for video data traffic, then the implementation is straightforward, but the reliability of SLA performance prediction deteriorates
Solution Approach 1:
The patent transitions from static Poisson modeling to dynamic Markovian Arrival Process modeling that can adapt to changing traffic conditions. The MAP model captures time-varying characteristics of video traffic through state-dependent transition rates, allowing the system to dynamically adjust its predictions based on observed traffic patterns, thereby improving reliability while maintaining implementation feasibility through standardized computational procedures.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
This application discloses an SLA performance prediction method and a related apparatus, and a device. In the method, a first network device obtains timestamps and packet lengths of a plurality of data packets; determines an arrival time interval of each of the plurality of data packets based on the timestamps of the plurality of data packets, where arrival time intervals of the plurality of data packets constitute a time interval sequence according to an arrival order; determines time domain feature parameters of the plurality of data packets based on a Markovian arrival process MAP model and the time interval sequence, where the time domain feature parameters are parameters of the MAP model; and determines packet length feature parameters based on the packet lengths of the plurality of data packets that have arrived, to predict SLA performance of the plurality of data packets based on the packet length feature parameters and the time domain feature parameters.