CLASE 17 02 26

Understanding Probabilistic Dimensioning in Network Traffic

Introduction to the Session

• The session begins with a brief introduction, indicating that the focus will be on dimensioning based on probabilities.

• The instructor confirms visibility of their screen and prepares to share resources related to the exam.

Basic Concepts of Demand Calculation

• The instructor discusses using exam resources for probabilistic dimensioning, emphasizing the calculation of gross demand based on simultaneity factors.

• Definitions are introduced: users for video calls (N), traffic for video calls (R), and similar definitions for navigation and educational streaming.

Estimating Total Demand

• Total demand is estimated by summing user numbers multiplied by their respective traffic units, resulting in 64.5 Mbps as total demand before applying simultaneity factors.

Application of Simultaneity Factor

• The concept of simultaneity factor (f), ranging from 0 to 1, is introduced as crucial for estimating simultaneous demand.

• Specific simultaneity factors are assigned: 25% for video calls, 40% for navigation, and 20% for streaming.

Classical Model vs. Variability in Factors

• A classical model is used to calculate simultaneous demand, yielding approximately 19.45 Mbps after applying the simultaneity factors.

• Discussion shifts towards potential issues if simultaneity factors vary or assumptions prove incorrect.

Importance of Empirical Data

• The instructor raises concerns about relying solely on literature-based simultaneity factors without empirical data from field studies.

• Emphasizes that urban network implementations require more accurate data than secondary sources can provide.

Statistical Approach to Simultaneity Factors

• Highlights the need for statistical analysis in determining reliable simultaneity factors rather than arbitrary values from literature.

• Stresses understanding past data patterns through statistics to improve accuracy in future estimations.

Understanding Bernoulli's Theorem in Telecommunications

Introduction to Bernoulli's Theorem

• The discussion begins with the introduction of Bernoulli's theorem, which is commonly used in telecommunications. It highlights the existence of two states.

Defining States and Probabilities

• Two states are defined: I (active) and I sub K (inactive), where 1 indicates an active state and 0 indicates inactivity. This framework is essential for analyzing user activity in telecommunications.

• The probability of a user being active is denoted as P, while the probability of inactivity is represented as 1 - P. These probabilities form the basis for understanding user behavior.

User Activity and Traffic Analysis

• Each service I has a corresponding number of users (n sub i). Traffic per user (R sub I) only occurs when users are active, emphasizing the need to analyze active users specifically.

• For each active user, there exists a probability P sub I that determines their activity level. If they are active, they consume R sub i megabits per second.

Demand Determination

• The total demand for traffic depends on both the number of users and their individual traffic consumption rates. This leads to defining T sub I as total traffic for each service.

• Total traffic can be calculated using T sub i = R sub i * X sub i, where X sub i represents the number of active users.

Summation and Extended Equations

• An extended equation for total traffic (T sub t) incorporates all users by summing up individual traffics: T sub t = Σ(R sub i * X sub i).

• This approach ensures that all potential users are considered, not just those currently active.

Understanding Mean Values in Probability

• A mean value or average helps summarize data efficiently; it can be derived from either individual assessments or probabilistic means.

• The mean value specific to Bernoulli variables is denoted as E, representing expected probabilities related to user activity.

Calculating Mean Probabilities

• To calculate this mean probabilistically: E = 1 * P + 0 * (1 - P), simplifying down to E = P.

• This establishes a foundation for further calculations regarding binomial distributions related to active users.

Understanding Bernoulli's Mean and Variance

Calculating Expected Users and Traffic

• The calculation of expected users involves multiplying the number of users (n_i) by their probability of being active (p_i). This is essential for understanding user engagement.

• The focus shifts from calculating the mean of active users to determining the mean traffic, which is more relevant for network analysis. The traffic can be expressed as a summation involving R_i and X_i.

Linear Expectation in Traffic Analysis

• The concept of linear expectation is introduced, where e is associated with X_y through equivalence. This relationship helps in simplifying calculations related to expected values.

• The formula derived indicates that the mean traffic (E_t) equals R_i multiplied by n_i and their respective probabilities, leading to an understanding of Bernoulli's mean.

Interpretation of Daily Consumption

• The equation provides insights into expected daily user consumption, representing the statistical average traffic based on user activation probabilities.

• A low or high probability affects the average traffic significantly; thus, having accurate statistical data allows for better network dimensioning based on Bernoulli's mean.

Understanding Variance in User Data

• Transitioning to variance, it’s explained that variance measures how much individual data points deviate from the average. A lower variance indicates consistency among user behaviors.

• An example illustrates that if a class has a mean score of 8 with a variance of 3, most scores will cluster around this average but may vary within three points.

Implications of High vs Low Variance

• In telecommunications, low variance in traffic patterns is preferable as it suggests uniformity in user behavior. High variance indicates diverse consumption patterns requiring robust network planning.

• If there’s significant variation in usage patterns (e.g., some users consuming much more than others), network capacity must be adjusted accordingly to handle fluctuations effectively.

Statistical Foundations and Formula Adjustments

• Discussion emphasizes that while averages provide insight into performance metrics, high variances necessitate deeper investigation into underlying causes such as teaching methods or user engagement strategies.

• The standard formula for variance is presented but noted that when dealing with probabilities rather than concrete data sets, adjustments are necessary for accurate representation.

Defining Bernoulli's Variance

• The original definition of variance involves squaring deviations from the mean. It’s crucial to clarify what constitutes "mean" within this context—specifically referring to statistical averages.

• Emphasis on correctly applying these concepts ensures clarity when interpreting results related to both probability distributions and real-world applications like network management.

Understanding Basic Definitions in Probability

Defining Key Concepts

• The speaker introduces the basic definition of a concept, indicating that it is foundational to the discussion.

• The mean (media) is defined as "mu" (μ), which will be equated to "e" for clarity and consistency throughout the explanation.

Variance Explanation

• The variance is introduced, specifically referring to "y sub k," with an emphasis on its calculation based on probabilities.

• When y sub k equals 1, the variance formula becomes 1 - p^2, highlighting the probability of user connectivity.

• Conversely, when y sub k equals 0, the variance simplifies to p^2, demonstrating how user activity affects variance calculations.

Pondering Results

• The speaker discusses weighting both results based on user activity probabilities: active users have a probability of "P," while inactive users have a probability of 1 - P.

• A combined equation is proposed for better understanding: P cdot (1 - p^2) plus (1 - P) cdot p^2.

Transitioning to Binomial Phase

Moving Towards Binomial Distribution

• The speaker emphasizes that they are calculating variance concerning user activity and must transition this into a binomial framework.

• Clarifications are made regarding Bernoulli's variance and its application within telecommunications models.

Constructing Binomial Models

• Introduction of variables where X_i represents the number of active users and Y_k denotes their collective probability of being active.

• Acknowledgment that if all users are active, their individual probabilities converge towards 1; conversely, if none are active, it approaches 0.

Summarizing User Activity Variance

Summation and Independence in Variance Calculation

• The total number of active users can be expressed as a summation from K = 1 to N (total existing users).

• It’s noted that since each user's activity is independent, the overall variance can be calculated by summing individual variances.

Final Variance Formula Derivation

• Conclusively, the overall variance for all active users can be represented as n_i cdot p_i cdot (1 - p_i), establishing a clear relationship between user count and their respective probabilities.

Understanding Variance in User Models

Introduction to Variance in Models

• The discussion begins with the concept of variance within a model, highlighting its significance in understanding user behavior.

• The complexity increases when users are not independent; this introduces challenges that may not be covered in the current cycle.

Implications of Dependent Users

• When users depend on each other, the model's structure changes significantly, necessitating different techniques for reuse and analysis.

• An example is proposed to illustrate how to apply these concepts practically, focusing on calculating mean and variance based on given data.

Calculating Mean and Variance

• The calculation of mean is introduced using specific variables (R sub i, n sub i, p sub i), emphasizing their roles in determining traffic metrics.

• A detailed breakdown of how to compute variance follows, stressing the importance of including all relevant factors such as user activity and probabilities.

Identifying Errors in Calculation

• The speaker acknowledges an oversight regarding missing traffic data necessary for accurate variance calculations.

• Clarification is provided that previous calculations pertained only to active users rather than overall traffic variance.

Finalizing Traffic Variance Calculations

• The correct formula for traffic variance is established: it incorporates both user activity and probability factors effectively.

• Concluding remarks highlight the average traffic output alongside its associated uncertainty, providing insights into network demands.

Traffic Analysis and Risk Assessment in Network Management

Understanding Average Traffic and Variance

• The average expected traffic is approximately 19.425 megabits, with a variation range between 14 and 24 megabits.

• The speaker emphasizes the importance of understanding how much traffic can vary and the associated risks when provisioning network resources.

• It is suggested that one should contract for the maximum expected traffic (24 megabits), rather than just the average, to ensure adequate capacity.

Statistical Concepts in Traffic Management

• Discussion on risk assessment involves recalling probability concepts, particularly percentiles, which help in understanding data distribution.

• A Gaussian distribution is referenced to illustrate where most traffic data will be concentrated around the mean (19.425 megabits).

Percentiles Explained

• The concept of percentiles is introduced; specifically, the 50th percentile represents the median value below which half of the data points fall.

• An example involving grades illustrates how percentiles work: being at the 50th percentile means you are performing better than half of your peers.

Implications of High Percentile Data

• If most data falls within the 50th percentile, it indicates that network traffic is around average levels.

• Conversely, if data falls within the 95th percentile, it suggests that network usage is nearing its maximum capacity (around 24 megabits).

Decision-Making Based on Data Patterns

• When evaluating network performance statistics, it's crucial to determine whether most data clusters around a specific percentile to inform resource allocation decisions.

• The speaker notes that historical patterns or similar scenarios must be analyzed when implementing new networks due to potential lack of existing data.

Importance of Historical Data in Provisioning

• Accurate estimations for average user connections and their probabilities are essential for determining necessary bandwidth provisions based on statistical studies conducted by organizations.

Understanding User Connection Probability

Key Concepts in User Connectivity

• The probability of all users connecting simultaneously or using the video call service is discussed, with values set at 0.4 for navigation and 0.3 for educational streaming. This indicates variability in user engagement based on service type.

Exam Preparation

• Students are reminded that this theoretical content will be included in their upcoming exam, emphasizing the importance of thorough study. The instructor encourages students to review the material carefully.

Presentation Logistics

• A discussion arises regarding sharing a link to a shared board for student presentations, highlighting communication issues among students about accessing shared resources. The professor expresses frustration over previous confirmations not being made by students.

Project Update on Antenna Location

Change of Location Due to Environmental Factors

• Students report a change in project locality due to unfavorable weather conditions affecting autonomous power supply to antennas; they have now selected Catacaos district in Piura as the new site for antenna installation.

Coverage Mapping

• A coverage map is presented showing no existing operator coverage in the area, indicating potential opportunities for service provision and infrastructure development within three populated centers: Noria, Vegamoro, and Morante.

Energy Supply Considerations

Energy Infrastructure Assessment

• Discussion focuses on energy supply challenges due to geographical dispersion; it’s noted that some areas lack guaranteed electrical service which complicates project implementation plans involving repeaters and energy sources like solar panels.

Temperature Impact on Equipment

• The average temperature range (25°C - 30°C) is mentioned as a factor influencing equipment performance and sustainability in Catacaos district's climate conditions during project execution discussions.

Technical Challenges with Signal Quality

Client Coverage Issues

• It’s acknowledged that certain geographic features hinder signal quality; approximately 100 houses are identified as potential clients but only 31 are currently simulated due to these limitations impacting network design strategies.

Sensitivity Adjustments Needed

• Technical difficulties arise concerning sensitivity levels required for effective signal reception; adjustments are suggested to improve performance metrics across various client setups while considering equipment sophistication needs per location specifics.

Radiation Pattern Adjustments

Optimization of Radiation Patterns

• Recommendations include adjusting radiation patterns used in simulations due to excessive energy output causing interference; this highlights ongoing efforts towards optimizing network configurations for better overall performance across different client scenarios.

Radio Mobile Configuration and Data Analysis

Antenna Height and Channel Width

• Discussion on the importance of maintaining the same channel width to avoid issues. Suggestion to elevate antenna height or adjust azimuth for better performance.

• Confirmation that adjustments made have improved the situation, indicating a successful troubleshooting process.

Video Resources for Radio Mobile

• Mention of finding a video tutorial on exporting units in Radio Mobile, emphasizing its utility for users.

• Explanation of converting KMZ files to KML format for easier data import into Radio Mobile, streamlining the workflow.

Practical Assignments and Python Integration

• Introduction of an assignment requiring students to create a traffic database using Python, with emphasis on quality work impacting grades.

• Clarification that statistical analysis should include at least 350 users to derive meaningful insights from data visualizations.

Understanding Traffic Patterns

• Description of how routers distribute megabits from providers to users through switches, highlighting individual consumption patterns.

• Importance of collecting data on user consumption patterns (average and variance), which aids in network planning based on observed behaviors.

Statistical Analysis in Network Design

• Discussion about avoiding confusion around terminology like "bandwidth," stressing accurate language use in professional contexts.

• Encouragement for students to apply AI and statistical methods learned in class when analyzing traffic patterns for their projects.

Group Work and Project Expectations

• Outline of project expectations where group collaboration is encouraged; grading will reflect effort and quality of analysis presented.

• Emphasis on detailed statistical analysis (mean, variance, standard deviation), crucial for understanding traffic engineering concepts.