Release Notes for Diffusion 3.2.0 --------------------------------- 1. Compiling Diffusion Starting with Diffusion 3.1, we use autoconf/automake to configure diffusion. There are several options that can be used with configure (for a complete list, please type ./configure --help). Relevant command line options for configure are: --with-ethernet // Enables wired Ethernet support in diffusion --with-wins2 // Enables support for the Sensoria WINSNG 2.0 radios --host=sh4-linux // Enables support for using the sh4 cross compiler // It assumes that the sh4 cross-compiler provided by // Sensoria is already installed and on the user's path. // The configure script will default to gcc/g++ if // sh4-linux-gcc and sh4-linux-g++ cannot be found. --prefix=destination_path // Enables the user to specify where the binaries will // be installed (default is the test directory inside // the distribution) So, compiling diffusion for the Sensoria nodes, should be as simple as: bash# ./configure --with-wins2 --host=sh4-linux bash# make bash# make install On the other hand, compiling for wired Ethernet on an x86 machine is: bash$ ./configure --with-ethernet bash$ make bash$ make install Please remember to remove the config.cache file when compiling for a different environment, as the file is not always deleted and may contain misleading information for the configure script. We tested compiling diffusion on Red Hat Linux versions 7.2, 7.3, 8.0, and 9.0. It also compiles and runs on other versions of Unix. 2. Before Running Diffusion (Sensoria Hardware) In order to use diffusion on the Sensoria nodes, you will have to have streamd running on your system. We assume streamd will export the device files /dev/rf/0/flink and /dev/rf/1/flink. Diffusion will not start if these files are not present. Starting with Sensoria software version 2.0, streamd is started automatically when nodes boot, so there is no need to start it manually for the two radios. In addition, diffusion expects the radios to be configured as base/remotes and clusters to be configured before it starts. Use /dev/rf/*/command to manually configure each radio with the desired topology (or use clusterd to set up a cluster automatically if a specific topology is not important). 3. Setting up Node Ids Diffusion uses a random ID for the each node. This behavior can be overridden by setting the 'node_addr' environment variable to the desired ID. Under bash, this can be done with: bash$ export node_addr=2 When diffusion is compiled to run on emsim (with the --with-emsim configure flag), it can get its node id directly from emsim's environment variables. Gear will also be able to get node location from emsim automatically. 4. Configuration File (Wired Ethernet Environment) Note that in the wired case, the file 'config.txt' (or an alternate configuration file, specified with the -f flag) should be present and contain a list of all directly connected nodes, one node in each line. If connected nodes are running on a port other than the default diffusion port (done by using the '-p' flag), this must be indicated in the configuration file as well. When diffusion initializes, it will display the list of connected nodes. Also, the configuration file can include the link quality to each neighbor. It's an integer from 0 to 100 and will indicate the probability of a message sent by your node to reach this neighbor. A sample configuration file is: 10.0.0.1 - 3400 - 100 10.2.4.1 - 30 10.3.12.4 - 85 In this case, we have three neighbors (10.0.0.1, 10.2.4.1 and 10.3.12.4), to which the probability of a packet getting lost is 0, 70% and 15% respectively. Note that the first neighbor (10.0.0.1) is running on the port 3400 (instead of the default diffusion port, 2000). Since the link quality is an integer from 0 to 100, and the port number has to be above 1024, diffusion can correctly guess the second argument's meaning if there are just 2 arguments in a config line. 5. Running Diffusion To run diffusion, simply start the 'filter_core' main process followed by the applications/filters desired. Please allow a few seconds for the filter_core to initialize before starting applications and filters. Be sure to at least start the two_phase_pull filter (where the two-phase pull and one-phase push algorithms are implemented). For debugging, the log filter can be used to print all messages arriving at the node. To test if things are working, start filter_core, two_phase_pull and ping_sender in a node. Then start filter_core, two_phase_pull and ping_receiver in another node. ping_receiver should receive ping_sender's messages after a few seconds. For example: bash$ filter_core & bash$ two_phase_pull & bash$ ping_receiver To increase the amount of debug messages, you can use the '-d debug_level' flag. A debug_level 10 will generate a lot of output. The default value is 1. Please note that if filter_core is run with the '-p port' flag, all other applications will also need it (i.e. two_phase_pull -p port, agent1 -p port, etc...). Also, for using the GEAR protocol, you will need to start gear as well, So, the list of processes to start is: bash$ filter_core & bash$ two_phase_pull & bash$ gear & bash$ your_application & --------------------------------- Fabio Silva Information Sciences Institute University of Southern California