The ONL NPR Tutorial

NPR Tutorial >> Examples

TOC

The Remote Laboratory Interface

Errata

• In some bandwidth charts below, the Y-axis label shows bytes/sec when the units were changed to Mb/s in the chart label dialogue box. This is due to a minor bug in the chart label dialogue box.

  Workaround: Using View => Change Y-axis Label in the monitoring window, sets the units for all of the plots in the chart.

Introduction

We present an example using two NPRs in a dumbbell configuration. Although one of the simplest configurations uses only one NPR, the two-NPR arrangement directly represents a popular configuration that appears in many textbooks and research papers. We will create a dumbbell topology and then send ping traffic from host n1p3 on the left to host n2p2 on the right through port 1.4 Then, some common problems/mistakes are listed. Finally, you will be asked to modify the example to support two ping flows instead of one with both flows going through port 1.4 but with the returning reply packets going through the link added between ports 2.4 and 1.1.

Some of the major features covered by this example are:

• Defining the (virtual) hardware configuration using the RLI and commiting to actual hardware
• Moving icons
• Setting up default routing
• Monitoring traffic bandwidth and packet rates
• Adjusting monitoring displays/charts

The example is given as a list of steps accompanied by screen shots. At the end is an example that tests your understanding of RLI basics.

A few comments before we begin:

• You can define a configuration of routers and hosts using the RLI without connecting to the ONL testbed.

  This approach of defining the experiment configuration while disconnected from the testbed will help you reduce setup errors by giving you time to reflect on your choices. Furthermore, it will improve resource usage efficiency during testbed busy periods.

• You should have already followed the directions given by the Getting Started link in sidebar of the ONL Web site.
• We have not included screen shots of logging into the ONL user host onlusr since there are a plethora of tools. But here are a few comments:
  • We use a password-free SSH approach described in Tutorial => Summary Information => Password-Free SSH.
  • The screen shots in the examples were captured while running Windows XP, and we used the PuTTY tool for enterring commands. Many of us actually use Linux or Windows XP with Xterms, but that is a personal choice.
• You may (or may not) encounter some problems with printing this web page because of the large number of images. We have not found any automated, reliable method to avoid vertical compression of images that fall on a page boundary.

Steps (Index)

Here are the major steps in this example:

1. Create the SSH tunnel required by the RLI
2. Start the RLI on your client host
3. Add two NPRs and four hosts to the experiment
4. Save the configuration
5. Make a reservation for 1 hour
6. Commit the experiment
7. Set up default routing at each port
8. Monitor some packet rates
9. Rescale the bandwidth display
10. Monitor some traffic bandwidths

Steps (In Detail)

1. Create the SSH tunnel required by the RLI

   See the Getting Started link in the sidebar of the ONL Web page. This step is not required if you do not plan to commit your experiment or make a resource reservation through the RLI.

2. Start the RLI on your client host

   Typically, this is either done through the command java -jar RLI.jar or by double-clicking on the Desktop RLI icon.
   Watch Out!!! If you are trying this before Fall 2008, you should use the RLInpr.jar file.

3. Add two NPRs and four hosts to the experiment

• Select Topology => Add NPR (clicking the center of the icon will show it it's logical name is NPR.1)

Note: We'll deal with the Commit Needed message at the bottom of the window later.



• Drag the icon away from the upper left corner of the window with the left mouse button while holding it down on the center of the icon.

• Repeat for the second NPR and move it to the right of the first NPR. (this is NPR.2)

We will now spin NPR.1 to the right and NPR.2 to the left to improve the appearance of the dumbbell when connecting them to each other and the four hosts. This is done with the spin handle (pink circle) attached to port 0.



• Select and hold on to the spin handle while rotating each of the NPR icons into place (see right).

Watch Out!!! A spin handle is picky about where you grab it. For best results, aim towards the southeast corner of the spin handle circle. It tells you that you have grabbed the right spot by turning itself bright red.
Also, a spin handle can only be moved to the five angular positions at the circular grab points along the outer perimeter of the icon.



• Select Topology => Add Host to get a host and drag it into position using the left mouse


• Select Topology => Add Link and drag from the host icon to the center of port 3. (Dragging from the port to the host also works.)

Note: The host name has changed from HST0 to n1p3 indicating that the network interface to the NPR will have a logical name of n1p3 (NPR 1, port 3).



• Repeat for the three hosts that will be attached with the network interfaces n1p2, n2p2, and n2p3.

The host icons are light blue and the links are dashed to indicate that no actual hardware has been acquired and initialized.



4. Save the Configuration

   Saving a configuration aids in recovery from misconfigurations or a testbed failure. It is good habit to adopt a discipline of periodically saving configurations as they evolve.


• Select File => Save As

A dialogue box will appear.



• The RLI assumes that your configuration files will be in the .onldir directory in your home directory.

The figure (right) shows that my .onldir directory already has lots of configuration files. Because these files are for the older NSP routers, I have created the npr subdirectory for all of my NPR configuration files.
The five icons at the top of the dialogue window allow you to:

• Change to the next directory up
• Change to your home directory
• Make a new directory in the current directory
• Show the current directory in brief
• Show the current directory in detail

Note: You can always get back to the .onldir directory by selecting the Home icon.



• The figure (right) shows that I have changed into my npr subdirectory and want to save the configuration as ex-rli.exp.


5. Make a Reservation for 1 hour

   This step is unnecessary if you already made a usable reservation earlier.


• Select File => Make Reservation.
• Fill in the reservation duration in the Make Reservation window.

  A dialogue box titled "Log In" will appear.



• Enter your User Name and Password in the dialogue box.


• A status message will appear in a Reservation popup window once the reservation has been processed.


6. Commit the experiment

• Select File => Commit


• The red Commit Needed box should disappear and the progress bar should show the text "Committing"

This binding process usually takes from one to four minutes.
Note: The longer times usually occur when there are lots of users trying to configure their experiments or if there are hardware problems.
What's Happening? The ONL testbed control software must now configure all of the components to support your description of the network. In this example, it must find two free NPRs and four free hosts and initialize them to their default settings. It also has to configure the switches to give you the connections needed for the dumbbell topology.


• You can observe the progress of the allocation and initialization process through the ONL Web page.

If the information is accepted by the ONL resource daemon, resource allocation and initialization will commence. The initialization of NPRs and hosts takes a few minutes. Because we use proactive rejuvenation (do standard initialization after an experiment terminates), resource intialization can appear to be almost instantaneous when no one else is using the testbed. Also, minor changes such as changing a routing table occur almost instanteously. You can observe the progress of the allocation and initialization process through the Testbed Status link in the sidebar of the ONL Web page.



• During the commit process dashed lines will change to solid ones; and the host icons will turn to a dark blue;


• After the commit is done, and the progress bar will show a "Commit Completed" message. Selecting the center of the host icon (the n1p3 host below) shows the control network interface name (onl052.arl.wustl.edu) and the internal data network interface IP address (192.168.1.64).
  Selecting the center of the NPR icon (NPR.1 above) shows its base IP address (192.168.1.0) and the name of the Control Processor for the NPR (nprcp09.arl.wustl.edu).
  

There are menu items associated with hosts and NPRs that will be covered in other examples.

7. Set up default routing at each port

What we have so far is four hosts connected through two routers that have default settings. But the default setting of an NPR includes the following:

1. An empty routing/forwarding table
2. An output link capacity of 1 Gbps on each outgoing link.
3. 64 datagram queues on each output port that are fed by a Stochastic Fair Queueing algorithm
4. Each datagram queue can hold 32,768 bytes.

Item 1 means that any packet coming into a router will be dropped because it doesn't know to which output port the packet should be sent. This step fixes that problem.


• Right click on the n1p3 icon to get name of the external interface (onl052.arl.wustl.edu in this example)
• Ssh to onl.arl.wustl.edu

You will actually end up on the ONL user host onlusr.arl.wustl.edu.

• From the host onlusr, ssh to the control/external interface of the host assigned to n1p3

The example shows that the control (external) network interface is onl052.arl.wustl.edu (onl052 is an alternate name).

• Ping n2p2 from the n1p3 host

(In the commands below, you will have to replace the host names such as onl052 with the names of your hosts.) This is what we have done so far:

client> ssh onl.arl.wustl.edu
onlusr> ssh onl052
onl052> ping n2p2

Watch Out!!! Each host has two network interfaces: 1) a control (external) network (e.g., onl052), and 2) a data (internal) network (e.g., n1p3). Each network interface has an IP address. The control network interface has an IP address of the form 10.0.0.X. The data network interface has an IP address of the form 192.168.X.Y where X is the NPR number and Y is something like 32, 48, 64, or 80. Usually, we want to send test traffic (e.g., ping traffic) over the data network. But when we SSH from one host to another, we usually want to do this over the control network.

Remember: You must first SSH to the user host onl.arl.wustl.edu before you can SSH to any of the hosts assigned to your experiment.

What's Happening? The ping command failed because the routing table is empty. Since the target of the ping command refers to the internal network interface n2p2, the host attempts to send the ping packets over the internal network interface which is associated with the name n1p3 and the IP address 192.168.1.64. This internal interface is connected to NPR.1, but NPR.1 will drop the packets because there will be no routing table entry matching the packets' destination IP address (192.168.2.48).



• Look at NPR.1's routing table at port 3 to verify that it is really empty:
  Select NPR.1/Port 3 => Configuration => Route Table


• Install default route table entries at all input ports
• Select Topology => Generate Default Routes
• Select File => Commit


• Look at NPR.1's routing table at port 3 to verify that it now contains a default routing table:
  Select NPR.1/Port 3 => Configuration => Route Table

Note: Since the ping packets are destined for n2p2 which has an IP address of 192.168.2.48, they will match the last entry in the routing table and be sent out port 4 on NPR.1.



• Notice that the ping command now succeeds starting at packet number 13.


8. Monitor some packet rates

• Select Add Monitor Display from the Monitoring window and type "pkts per sec" into the popup window.

  A new window titled "pkts per sec" will appear.



• Select Monitoring => Set Default Polling Rate and type "0.25" into the popup window.

Note: In most situations, polling once per second will show sufficient detail. If we polled once per second with ping traffic (one packet per second), the bandwidth chart would appear as a horizontal line. But by polling once every 0.25 seconds, the bandwidth chart will show spikes occuring once per second.



• In the pkts per sec window: Select Add Parameter

Strictly speaking, this step is unnecessary when there is only one monitoring chart since the RLI is smart enough to realize which chart is being referenced when there is only one monitoring chart.



• In the RLI window: Select NPR.1/Port 3 => Monitoring => RXPKT

Note: The number of of packets coming into NPR.1's port 3 will be monitored every one second. For ping packets, we expect to see one packet very second.
Note: The TXPKT menu option measures the number of packets transmitted from an output port. The RXBYTE and TXBYTE options count in units proportional to the number of bytes (e.g, Mbps).



• In the Add Parameter window: Check the rate checkbox so that the number of packets per second will be displayed. The alternative is to display the absolute value of the packet counter.

The label "RXPKT" appears in the legend and a line appears.
Note: You do not need to do a commit to do any charting operations.



• In the pkts per sec monitoring window: Double click the RXPKT label in the legend, and change the name to "RXPKT 1.3" to distinguish it from other similar labels. The notation 1.3 is shorthand for NPR.1/port 3.
  Note: Another popular option is to change the line type from solid to dashed by checking the Dashed Line checkbox.


• Repeat the above steps so that you monitor the five packet rates shown to the right (steps not shown). The display is for the case when the command ping n2p2 is run on from the n1p3 host.

What's Happening? By default, the ping command sends one packet per second. But the display shows a packet rate of about 4 packets per second. This is because we are polling at a 0.25 second period. When the packet counter does increase, it increases by one in a 0.25 second interval which is displayed as a rate value of 4 pps = 1 pkt/0.25 second.

9. Rescale the bandwidth display

The appearance of the previous chart can be modified to reveal more of the ping-like behavior: bandwidth spikes appearing approximately once per second. The charting features of the RLI allow the user to focus on a portion of a chart by zooming in and out along both the X- and Y-axes.


• Left-clicking the X-axis at point X zooms in so that the rightmost X-value is X and also freezes the chart so that new values are not displayed. Dragging the X-axis slider to the far right resumes the display of new values.

Note: There are four spikes within every four second interval.
Remember: A particular line in the chart can be highlighted by passing the cursor over the legend label.



• Left-clicking the y-axis at point Y zooms in so that the topmost y-value is Y.


• You can show the approximate X- and Y-coordinates within a chart by selecting View => Show Values and then left clicking the display point.


• (Not Shown) Zooming in on a window can also be done through the menu item View => Zoom In From Selection.
• You can zoom out by left-clicking on the arrowhead of the X-axis or the Y-axis.
10. Monitor some traffic bandwidths

    Follow the same steps to create a new chart which shows the traffic bandwidths in Mb/s (with title bandwidth).

    
    • RLI Window: Select Add Monitor Display and type "bandwidth" into the popup window.

      A new window titled "bandwidth" will appear.

    • bandwidth window: Select Add Parameter
    • RLI window:
      Select Port 1.3 => Monitoring => RXBYTE Parameter and check the rate box.

    Note: The default polling interval will still be the 0.25 second that you globally set earlier.

    
    
    • bandwidth window: Open up the RXBYTE dialogue box by clicking the RXBYTE label, and select Mb/s units.
    
    
    • Repeat the above steps for TXBYTE 1.4, RXBYTE 2.1, TXBYTE 2.2, and TXBYTE 1.3.

Monitoring Menus

Most monitoring window menu items are intuitive. Below is a brief summary of the menus. An example of the Add Formula item is given in the example on filters.

• Options Menu
• Save to File ...: Create a configuration file that contains only the monitoring chart (i.e., exclude the network configuration).
• Change Title: Change the title of the chart.
• Stop: Stop the monitoring associated with the window.


• Parameter Menu
• Add parameter: Start adding monitoring items to this chart.
• Remove parameter: Remove a monitoring parameter.
• Add Formula: Arithmetically combine several time series into one chart.


• View Menu
• Show Values: Show the X- and Y-coordinates of a point in a chart.
• Reset View: Reset the y-axis so that the data will fit vertically.
• Change Y-axis label: Change the label of the Y-axis.
• Set Log Scale: Change the Y-axis to use a log scale.
• Zoom In From Selection: Define a new view window by dragging along the diagonal of a view box.

Common Problems/Mistakes

Here are some common problems/mistakes:

• Can't SSH to an ONL host

  You must SSH first to the onl user host onl.arl.wustl.edu, and then to one of our ONL hosts.
  Why? We have a firewall that disallows any direct connections to the ONL hosts assigned to your experiment.

• Can't ping an ONL host

  Reason 1: You can only send traffic through the internal network (connecting your ONL hosts) from a node connected to the internal network. You can not ping the internal network interface from the onl user host onlusr.arl.wustl.edu.
  Reason 2: You did not properly set up the routing tables. Remember that by default all routing tables start off empty.

• Only one line shows up on my 4-line chart

  If the lines fall on top of each other, you can highlight the others by passing the cursor over the lines label in the legend.

• Some lines on my monitoring chart are not shown

  Reason 1: Try the View => Reset view menu item and then check that the Y-axis values are reasonable. Reason 2: If you mix monitoring of rates with absolute values, the ones with absolute values may fall outside the window. You probably want all of them to be either rates or absolute values but not a mix of the two.

Test Your Understanding

Construct the dumbbell configuration shown right that will support two ping flows: 1) n1p3-n2p3, and 2) n1p2-n2p2. The ICMP echo request packets in the forward direction should take the northern route (go out port 1.4 to port 2.1), but the returning ICMP echo reply packets in the reverse direction should take the southern route (go out port 2.4 to port 1.1). How does the configuration file for this example differ from the one earlier?

Construct a pkts per sec packet rate chart with the packet rates shown to the right. Note that the RXPKT 1.1 plot is a dashed line while all of the other ones are solid lines. Can you explain the following features in the chart:

• The maximum value in the RXPKT 1.1 line (dashed brown) is about 8 packets per second.
• The maximum value in the RXPKT 2.1 line (solid green) is also about 8 packets per second.
• All other maximums are about 4 packets per second.

NPR Tutorial >> Examples	TOC