The SPP Tutorial

From ARL Wiki
Revision as of 23:16, 14 March 2010 by Ken Wong (talk | contribs)
Jump to navigationJump to search

Template:Under Construction

Introduction

This tutorial page and The Hello GPE World Tutorial page are designed to give you simple, step-by-step usage examples that will allow you to get started quickly. They also contain links to a number of auxilliary pages that will allow you to try more complex examples. Two other wiki pages complete the tutorial: Using the IPv4 Code Option and Developing a New Code Option.

You don't have to read all four pages before using the SPPs. But you can't do much unless you read this page and The Hello GPE World Tutorial page. But these two pages (like the other two) are designed so that you can get hands-on experience by following the prescribed steps.

(Note: Scan the Internet Scale Overlay Hosting page if you have not already done so paying particular attention to the parts describing the role of the GPE (General-purpose Processing Engine) and the NPE (Network Processor Engine). Keep in mind that in a high-speed implementation, the SPP would be configured so that the NPE handled most data packets (the fastpath) leaving the GPE to handle control and exception packets (the slowpath).)

This page will walk you through the preliminary steps required to get setup to use an SPP; e.g., getting an SPP account and adding SPP nodes to your slice.

The Hello GPE World Tutorial will show you how to use the GPE to create a GPE-only prototype. It will show you how to get information about interfaces; how to create slowpath endpoints for communicating with a process running on a GPE; and how to make a resource reservation. It will show you how to write a simple packet-processing program that will run on a GPE. This program can be extended to create a GPE-only prototype of a simple slice.

By this point, you should be able to do simple SPP experiments that will help you develop your own GPE-only slice software. Although this GPE-only software can not take full advantage of the SPP because it doesn't use an NPE, most users will not be writing code for NPEs because programming an NPE requires knowledge about IXP network processor programming and the SPP substrate layer. In practice, we anticipate that most developers will write a GPE-only version of their slice and work with an advanced SPP developer to transform the performance-critical parts of their GPE-only code to run on the NPE. That is, most developers will write a GPE-only program which captures the logic of their slice and pass this design to an advanced SPP developer who will create a high-speed fastpath (NPE-GPE) version.

Using the IPv4 Code Option describes how to use the IPv4 code option which uses an NPE to forward IPv4 packets and a GPE to process packets terminating at the router (e.g., local delivery and exceptions). This page will walk you through an example of how to use the IPv4 code option. It explains how to create fastpath endpoints, meta-interfaces, packet queues and IPv4 filters. You can directly use all of these concepts in developing your own slice except for IPv4 filters. However, your code option will need to use filters to direct incoming packets to packet queues. Seeing how it is done for the IPv4 code option will give you a better understanding of how to design filters for your own slice.

Developing a New Code Option walks you through an example of how to develop a simple code option. It shows you how to build a GPE-only slice for a simple protocol and discusses how that code might be transformed to use both an NPE and a GPE.

Tutorial Conventions

We show blocks of Unix commands and their output in monospaced typeface indented by four spaces. When a command and its output are shown together, the command is distinguished from its output by preceding the command with the prompt characters "> " (greater than, space). In the example,

    > ls
    foo    myfile    yourfile

ls is the command, and its output is shown without the preceding prompt characters.

User-specific command arguments will be shown as fully capitalized words. For example, in the command:

    ssh YOUR_SLICE@64.57.23.210

YOUR_SLICE should be replaced by the user's slice name.

The following command block shows two forms of commentary:

    ssh-add -v        # -v outputs debug info and is optional
    ... You will be asked for your passphrase ...

that only provides guidance. You are not suppose to key in these comments.

When a command sequence involves more than one host, the hosts are distinguished by the command prompts. For example,

    host> ssh YOUR_SLICE@64.57.23.210    # command entered at myhost
    SLC>  ls                             # command entered at SLC GPE

shows the first command entered while logged into myhost and the second command entered after logging into a Salt Lake City GPE.

Requirements

This tutorial assumes that you will be using a Linux operating system and have access to standard Unix tools. In particular, you will need:

  1. An environment where you can compile code for a Linux operating system
  2. An OpenSSH executable
  3. The latest version of Java to run some monitoring software and traffic generators

Compiling

You can use almost any Intel host running a recent version of Linux to compile code for the GPE. But if you think this is an issue, you can always try to match the version of Linux and glibc on a GPE. You can get this information after you ssh to a GPE (described later in The Hello GPE World Tutorial) by entering:

     uname -sr          # get version of Linux
     ldd --version      # get version of glibc

OpenSSH Executable

OpenSSH is a suite of open source utilities using the ssh protocol. Although you don't have to specifically use the OpenSSH suite, the tutorial examples assume that the ssh tools are from OpenSSH.

The Java Run-time Environment (JRE)

Most of the tutorial can be done without the JRE. But it will be helpful in the IPv4 Code Option Tutorial.

First, determine if you already have JRE 1.6.0 or higher. Enter the following in a command-line window: java -version. If the response says that you are already running some version that is atleast 1.6.0, you can skip this step.

If you need to install the JRE, go to the Java download page http://java.com/en/download/manual.jsp and download the latest JRE for your platform by following the instructions posted at the Sun site. Note: You only need to download the JRE (less than 20 MB), not the entire JDK (Java Development Kit).

MAC Users

Follow the instructions at Java Run-time Environment for the MAC OS.

Getting Started

You must get an SPP account before you can use the SPPs. The procedure is similar to the one used in PlanetLab except that SPP slices are administered through a special PLC (PlanetLab Central) site at http://drn06.arl.wustl.edu/. But the following PlanetLab procedure still holds:

  • A Principle Investigator (PI) must create a slice for you.
  • The PI must assign you to that slice.
  • Then, you can add SPP nodes to your slice.

In a typical scenario, you (as an ordinary PlanetLab user) will use both normal PlanetLab nodes and SPP nodes. This requires that you get assigned both a normal PlanetLab slice and an SPP slice. Once you have been assigned to an SPP slice, you can log into the GPE of an SPP slice like any other PlanetLab node using ssh and develop a socket program as you would for any other PlanetLab node except for a few small differences:

  • An SPP has multiple network interfaces if you want to use more than one interface.
  • You must configure the SPP before you can use it.

The configuration process is fairly straightforward since a user typically customizes an existing configuration script to do the configuration.

Creating an SPP Account

To create an SPP account, begin by using your browser to go to the SPP PlanetLab web site http:drn06.arl.wustl.edu. Select the Create an account link, and a registration form will appear. Fill out the registration form and select the Register button at the bottom of the page. Your email address will be used as your SPP user name. So, make sure that you enter an address where you actually read email. After your account is created (which can take 24 hours or longer if you try to get an account on a weekend or holiday), you will be notified by email.

You will need to use ssh to log into to SPP nodes. So, while you are waiting for this notification, you can create your RSA key pair if you haven't already done so. You shouldn't need to do this if you already have a PlanetLab account because you can just use your PlanetLab RSA key pair. We assume below that you will be storing your RSA key pair in the standard place ~/.ssh/id_rsa (private key) and ~/.ssh/id_rsa.pub (public key). Use the ssh-keygen command to generate the RSA key pair:

    $ ssh-keygen -t rsa -f ~/.ssh/id_rsa
    Generating public/private rsa key pair.
    Enter passphrase (empty for no passphrase): ENTER SOME PASSPHRASE
    Enter same passphrase again: ENTER THE PASSPHRASE AGAIN

You will need to upload your public key to the SPP PlanetLab database after you receive email notification that your SPP account has been created. After you receive the notification:

  • Go to the SPP PlanetLab Central web site http:drn06.arl.wustl.edu.
  • Log into the web site.
  • In the navigation box on the left, click “My Account.”
  • In the “Keys” section, choose “Manage Keys”, and select “Browse” and navigate to your public key file (i.e. ~/.ssh/id_rsa.pub).
  • Check Upload. (sidebar "Users => My Account")

You should see XXXXX.

>>>>> ABOVE IS INCOMPLETE. WHAT ABOUT SITE AND ROLES? <<<<<


>>> What should PIs and Technical Leads do??? <<<

Adding SPP Nodes to Your SPP Slice

>>> SOMETHING GOES HERE <<<