Assignment Instructions:
For this assignment you will analyze the performance of the HTTP and Ethernet protocols with Riverbed Modeler in a LAN depending on the number of client nodes.
Create a scenario with 6 nodes connected to an Ethernet switch (ethernet32_switch node model) via 100 Mbps links. One of the nodes is an ethernet_server, the others are ethernet_wkstn nodes.
Leave the default parameters for the client nodes unchanged.
Configure the HTTP application configuration to have the following parameters:
• page interarrival time: exponential with a mean of 20 s
• in the page properties, set the page to have one object of size (constant, 10000 bytes) and 10 other objects of size (uniform, with range 5000-10000 bytes).
Leave the other application configurations to defaults. Create application definitions for email, FTP, and Database, all with the 'High Load' parameters (predefined).
Create a user Profile Configuration that has all the applications mentioned above, running them simultaneously beginning at a time distributed uniformly between 120 and 140 s from the simulation start.
Create the necessary scenarios (with 'Duplicate Scenarios') to evaluate the following performance metrics for a number of client computers ranging from 5 to 30 in increments of 5:
• HTTP page response time (global)
• HTTP object response time (global)
• HTTP Traffic received (global)
• Ethernet Delay (global)
Use copy-paste to add 5 client nodes for each new scenario. Name the scenarios in a meaningful way (e.g. eth-lan-5 for the scenario with 5 clients, eth-lan-10 for 10 clients, etc.).
Run all scenarios with 30 minutes simulation time each.
The statistics you work with involve the CDF metric. CDF means Cumulative Distribution Function and is defined as follows: for a random variable X (e.g. packet delay) the CDF F(x) is defined as the probability that the random variable is less than x:
F(x) = P(X < x)
Example: the Ethernet MAC access delay is a random variable. The probability that the MAC access delay for all nodes in the network (the "global metric") is less than 0.002 s could be 95% for some scenario. With math, this looks like F(0.002) = P(delay < 0.002s) = 0.95. This also means that 95% of all packets get a MAC access delay of less than 0.002 seconds.
The CDF is always an increasing function, with values between 0 and 1 because it measures probability. For most networking stats, the CDF is a continuous monotonous increasing function. For analyzing protocol or network performance, the CDF is most useful when looking at delay metrics, such as packet delay or jitter.
For data rates (or traffic received (bits) per second) the most common metric is the average (mean). The minmum and maximum are also useful for more detailed insight.
The assignment:
Create a report in Microsoft Word (.doc) format with the following items:
1 image (or screenshot) of the network topology (the scenario window) with 5 clients
2 image (or screenshot) of the network topology (the scenario window) with 30 clients
3 one image (or screenshot) for each of the above metrics being displayed 'overlayed' for all scenarios (use 'Compare Results' menu command):
3.1 HTTP page response time (global) - CDF
3.2 HTTP object response time (global) - CDF
3.3 HTTP Traffic received (global) - moving average with a window of 30 s
3.4 Ethernet Delay (global) - CDF
4 One horizontal bar graph for each performance metric in average value showing its dependency on the number of clients in the LAN. Show the average delay, not the average of CDF. You can export the data from a graph using right-click + Export Data to Spreadsheet. These graphs are easiest done in Excel. The horizontal axis shows the number of clients (5, 10, 15,...,30) and the vertical axis shows the corresponding metric (average!). The graphs need to be properly labeled and captioned.
5 One paragraph where you summarize your conclusions on the performance metrics dependency on client count.