2. CSE 5231 - Dr. Aydeger - Network Performance Metrics

Course Information

• Course Name: CSE 5231 Computer Networks

• Instructor: Dr. Abdullah Aydeger

• Location: L3Harris Center for Science and Engineering #326

• Email: aaydeger@fit.edu

• Department: Electrical Engineering and Computer Science

Network Performance Metrics

• Overview of key performance metrics used to evaluate network performance.

• Common metrics include:

  • Bandwidth

  • Throughput

  • Delay

  • Packet Loss

Network Protocols Overview

Layered Architecture

• Application Layer: End-user applications (e.g., web browsers).

• Transport Layer: Manages data transfer from end-to-end (TCP/UDP protocols).

• Internet Layer: Responsible for routing data from the source to the destination (IP protocol).

• Link (Physical) Layer: Physical media carrying the data (e.g., cables, wireless).

Data Communication Via Networks

• Application Data: User-generated data inputs.

• Transport Layer: Adds headers (TCP or UDP).

• Internet Layer: Adds IP headers or ICMP.

• Link Layer: Adds error correction and other necessary information.

TCP/IP Protocol Stack

• The four layers of the TCP/IP reference model enhance:

  • Modularity

  • Flexibility

  • Reliability

Data Handling in Layers

• Application Layer: Entire files (images, documents) considered as messages.

• Transport Layer: Divides messages into segments matching packet size limits.

• Network Layer: Refers to packets, which include data and headers.

• Data-link Layer: Refers to frames that include packets and additional headers.

Key Network Performance Metrics

Bandwidth

• Definition: Maximum bits-per-second a communication link can support.

• Example: 1 Mbps = 1 x 10^6 bits/second.

• Relationship to Transmission Time: Smaller bandwidth allows for more data in a given timeframe.

Throughput

• Definition: Actual data transfer performance (measured in bits/sec or packets/sec).

• Factors: Influenced by input rate and protocol overhead.

Bandwidth vs. Throughput

• Bandwidth: Represents the capacity of the communication link, ignoring delays.

• Throughput: Actual performance output, typically lower than bandwidth in real scenarios.

• Utilization: Ratio of throughput to bandwidth.

Packet Delay and Loss

Delay Sources

• Queueing Delay: Time packets spend waiting in router buffers.

• Packet Loss: Occurs when buffers fill up, leading to dropped packets.

Components of Delay

• Processing Delay (dproc): Time taken by routers to check packet errors and determine output link.

• Queueing Delay (dqueue): Time spent waiting for transmission.

• Transmission Delay (dtrans): Calculated as dtrans = L/R (where L = packet length and R = link transmission rate).

• Propagation Delay (dprop): Calculated as dprop = d/s (where d = link length and s = propagation speed).

Calculating Packet Delay

• Analogy: Cars waiting at a toll booth symbolize performance at a network link.

• Total Delay: Latency = Propagation Delay + Transmission Delay + Queueing Delay.

Jitter

• Description: Variability in packet latency; critical for applications like audio/video conferencing.

• Application Sensitivity: Different applications may prioritize bandwidth or latency differently.

Real Internet Measurement Tools

Traceroute

• Measures Internet delay and loss by sending packets to routers and receiving responses to determine the time taken for each hop.

Example: Ping Command

• Demonstrates the round-trip time for packets sent to a specific destination.

Packet Loss

• Causes: Arises when incoming packets exceed the buffer capacity at a network link.

Throughput Details

• Instantaneous vs. Average Throughput: Instantaneous rate is the current transmission rate, while average is measured over time.

• Bottleneck Link: The link in the path that limits overall throughput.

Network Scenarios

• Understanding how throughput is constrained by various components of the network and how it is calculated (e.g., min(Rc, Rs, R/10)).

Discussion Topics

• Critical delays affecting daily web traffic.

• Strategies for sending optimal data amounts without causing congestion.

• Calculating delays and throughput in practical scenarios (like the example of Hosts A and B).

Conclusion and Q&A

• Instructor Contact: Dr. Abdullah Aydeger available for questions.