nfn-scala is a NFN service layer implementation written in Scala, interfacing and depending on CCN-Lite.
This project provides an asynchronous compute environment, an CCN and NFN client library, a DSL to write lambda calculus programs.
It can also be used to run CCN-Lite directly from Scala and to create a network topology.
Installation is fairly straight forward, and boils down to an available Java JDK 7, sbt 0.13.5 and CCN-Lite.
If you want to install the JVM or sbt you could also test it out within a Docker container.
For a tutorial on that, refer to the docker tutorial in CCN-lite.
sudo apt-get install openjdk-7-jdk
- Follow instructions to install sbt
JDK 7 should be available, otherwise it can be downloaded directly from oracle.
To install sbt you can use homebrew (
brew install sbt).
To install ccn-lite and its dependencies follow the information here.
nfn-scala uses the commandline utilities of ccn-lite, therefore it needs to have the compiled utilities available.
There are two possibilities, either you have CCNL_HOME set to your custom installation or you clone the repository with
--recursive and use the submodule in
In general the latter is the easier option, because the version of ccn-lite is always compatible with your nfn-scala.
To compile ccn-lite (either in the submodule or, if not checked out,
$CCNL_HOME), you can type
Alternatively you can also build it manually, make sure that
USE_NFN=1 is set.
Checkout or download nfn-scala and swap to the nfn-scala root directory.
You can now use sbt to compile the project (e.g.
sbt compile) or you can use
This command will build the jarfile
./target/scala-2.10/nfn-assembly-0.1-SNAPSHOT.jar with all dependency (even Scala itself!)
which makes it easier to deploy it because you do not have to install sbt or scala (only JVM).
If you want to use IntelliJ or eclipse you can use the sbt tasks
For a detailed tutorial on both running CCN-Lite as well as nfn-scala go to the tutorial.
In the following two very basic ways to use nfn-scala.
Running a test project
There are some runnables classes in the project. To see them all you can simply use
sbt run and you will presented a list with everything available.
Choose for example .
runnables.evaluation.PandocApp which starts a nfn environment, sends a predefined request and prints the result.
Starting a standalone compute server
There is a small startscript/program to start and configure a single nfn-node.
This time we are going to build a jar-file containing everything (even scala) with
sbt assembly. This jar can be deployed on any JVCM.
To run it you can either use
scala ./target/scala-2.10/nfn-assembly-0.1-SNAPSHOT.jar -h (make sure you have the correct Scala version) or
java -jar /target/scala-2.10/nfn-assembly-0.1-SNAPSHOT.jar -h.
Running the above will print a help message. As you can see there are several options. The most important setting is if you want to start ccn-lite yourself or if you want nfn-scala to start it internally.
You do not have to worry about faces, they are setup automatically.
The prefix is the name under which all content and services are published.
If you want to use the default values (which starts CCN-Lite internally), you can run it with
sbt run /nfn/testnode.
To test if it everything works, use send the following two interests to CCN-Lite:
ccn-lite-peek -u 127.0.0.1/9000 "/nfn/testnode/docs/tiny_md" | ccn-lite-pktdump ccn-lite-peek -u 127.0.0.1/9000 "" "add 1 2" | ccn-lite-pktdump ccn-lite-peek -u 127.0.0.1/9000 "" "/nfn/testnode/nfn_service_WordCount 'this string contains 5 words'" | ccn-lite-pktdump
<!--- ## Visualization To replay and visualize the most recently run NFN program, change to the directory `./omnetreplay`. An installation of [OMNeT++](http://www.omnetpp.org) is required (we used Version 4.4.1, but other versions should work as well). Now you should be able to run the `make.sh` script which compiles and runs everything. From then on the simulation can be directly started with `./omentreplay`. -->
- CCN-Lite command line interface is slow, but it is convenient and currently helps to test CCN-Lite itself. For performance, the JNI interface (or JNA) should be reintroduced.
- Compute Server currently only does exact match on names and not longest prefix match.
This can result in issues, for example when computing on data from the testbed, which always adds a version number.
Since there is no implementaiton for a version number in CCN-Lite or nfn-scala, exact matches in nfn-scala will fail.
- Timeouts are not fully implemented (especially when using NACKS).