IcarusEdgeSim is a Python-based discrete-event simulator for evaluating the performance of both networks with both cache and compute resources based on the Icarus simulator (https://github.com/icarus-sim/icarus) for ICN caching networks.
Icarus is not bound to any specific ICN or edge-computing architecture. Its design allows users to implement and evalute new caching and computing policies or routing strategies with few lines of code.
The code is in general self-explanatory. Below are some notes on how to start experimenting with IcarusEdgeSim.
You can find sample configurations in the /examples folder.
A typical configuration contains information on the scenario to be executed. A general scenario includes
a network topology, workload, placement of caches, placement of data producers, and most importantly
a strategy. A strategy determines how requests are processed (i.e., routed from source to resources and possibly back),
a workload determines the arrival rate and placement of initial requests.
Example strategies are located under models/strategy/. The service.py in the examples contains strategies for edge-computing implemented for the following papers:
- In Store Edge Networked Data (SEND): A Data and Performance Driven Edge Storage Framework
- In Uncoordinated placement for edge-clouds
- In Fogspot: Spot pricing for application provisioning in edge/fog computing
These configurations include placement of caches and content (see ./scenarios/cacheplacement.py and ./scenarios/contentplacement.py), total storage budget (in number of content), total number of contents. A cache model comprises policies that manage how content is replaces under ./models/cache/ containing replacement policies. The placement and retrieval of content to/from caches at run-time is determined by a strategy.
These configurations include placement of storage and content in storage (as in cache - see ./scenarios/contentplacement.py), total storage budget (in Bytes), total number of contents. A cache model comprises policies that manage how content is replaces under ./models/cache/ containing replacement policies. The placement and retrieval of content to/from caches at run-time is determined by a strategy.
These configurations resemble caching configurations and include number of functions, total computation budget (in number of CPU cores) and computation placement. The placement setting locates a set of computation spots, i.e., Cloudlets, (see ./models/service/compSpoy.py) which comprises a number of CPU cores and VMs or containeras, in the topology. Each function is associated with a service rate, i.e., processing time per input data chunk.
A number of policies dictate how computation spots are managed. An admission policy determines how a node decides whether or not to accept a request for execution. A scheduling policy determines the order of execution for admitted requests. A strategy determines the placement of functions (mapping of VMs to function) for each computation spot.
The topology determines the connectivity of the network. Available topologies include measurement-based ones such as Rocketfuel (please see /resources/topologies).
This document explains how to configure and run the simulator.
SEND NOTE: Some more, Repo-(EDR-)specific versions of the previously implemented topologies are also available.
Before using the simulator, you need to install all required dependencies.
If you use Ubuntu (version 13.10+) you can run the script ubuntusetup.sh
located in the scripts folder which will take of installing all the
dependencies. To run it, executes the following commands
$ cd <YOUR ICARUSEDGESIM FOLDER>
$ sh scripts/ubuntusetup.sh
The script, after being launched, will ask you for superuser password.
Finally, it is advisable to add IcarusEdgeSim path to the PYTHONPATH environment variable. This makes it possible to launch IcarusEdgeSim from outside the IcarusEdgeSim root directory or call IcarusEdgeSim APIs from other programs:
$ cd <YOUR ICARUSEDGESIM FOLDER>
$ export PYTHONPATH=`pwd`:$PYTHONPATH
Note however that setting the PYTHONPATH this way does not persist across reboots. To make it persist you should add the export instruction to a script that your machine executes at boot or login time, e.g. .bashrc (if you use Bash).
If you have other operating systems, you can install all dependencies manually.
IcarusEdgeSim dependencies are:
- Python interpreter (2.7.x): you can either download it from the Python website or, possibly, from the package manager of your operating system.
- The following Python packages:
- numpy (version 1.6 onwards)
- scipy (version 0.12 onwards)
- matplotlib (version 1.2 onwards)
- networkx (version 1.6 onwards)
- fnss (version 0.5.1 onwards)
All these packages can be installed using either easy_install or pip utilities.
If you use pip run:
$ pip install numpy scipy matplotlib networkx fnss
If you use easy_install run:
$ easy_install numpy scipy matplotlib networkx fnss
You may need to run pip or easy_install as superuser. The installation of these packages, especially numpy and scipy may also require to install additional libraries.
You can also run IcarusEdgeSim within a virtual machine. This repository contains scripts and documentation to set up a virtual machine with IcarusEdgeSim and all its dependencies.
You can download a stable release in a zip or tar.gz format using the links below.
Latest version:
Older versions:
- Version 0.5.0: [zip] [tar.gz]
- Version 0.4.0: [zip] [tar.gz]
- Version 0.3.0: [zip] [tar.gz]
- Version 0.2.1: [zip] [tar.gz]
- Version 0.2: [zip] [tar.gz]
- Version 0.1.1: [zip] [tar.gz]
- Version 0.1.0: [zip] [tar.gz]
You can also get the development branch from the Github repository using Git. Just open a shell, cd to the directory where you want to download the simulator and type:
$ git clone https://github.com/icarus-sim/icarus.git
To use IcarusEdgeSim with the currently implemented topologies and models of caching and computing policies and strategies you need to do the following.
First, create a configuration file with all the desired parameters of your
simulation. You can modify the file config.py, which is a well documented
example configuration. You can even use the configuration file as it is just
to get started. Alternatively, have a look at the examples folder which
contains examples of configuration files for various use cases.
SEND NOTE: Note that this distribution of icarus is modified to run with repositories, thus the normal configuration needs to be changed in order for the whole system to work appropriately.
The configurations in the examples/repo-mgmt folder all work well and were used for the evaluation of the SEND system, as per the second cited paper, below. If you want to use the original features of icarus, please git clone and develop based on the above git repository or the most appropriate distribution for you.
Second, run Icarus by running the script icarus.py using the following syntax
$ python icarus.py --results RESULTS_FILE CONF_FILE
where:
RESULTS_FILEis the pickle file in which results will be saved,CONF_FILEis the configuration file
Example usage could be:
$ python icarus.py --results results.pickle config.py
SEND NOTE: Alternatively, there are bash scripts you can check, edit and/or run directly from the examples folder. These should also help you with any implementation of icarus or further development you may want to make for this distribution.
After saveing the results in pickle format you can extract them in a human
readable format using the printresults.py script from the scripts folder. Example usage could be:
$ python scripts/printresults.py results.pickle > results.txt
SEND NOTE: The new results format is available in just text format, strictly for statistical readings; for the purpose of academic and research implementations, evaluations and development.
Icarus also provides a set of helper functions for plotting results. Have a look at the examples
folder for plot examples.
By executing the steps illustrated above it is possible to run simulations using the topologies, cache policies, strategies and result collectors readily available on Icarus. Icarus makes it easy to implement new models to use in simulations.
To implement new models, please refer to the description of the original Icarus simulator provided in this paper:
L.Saino, I. Psaras and G. Pavlou, Icarus: a Caching Simulator for Information Centric Networking (ICN), in Proc. of SIMUTOOLS'14, Lisbon, Portugal, March 2014. [PDF], [Slides], [BibTex]
Further implementations and ideas, provided for the purpose of the current version of code were developed in:
A.-C. Nicolaescu, S. Mastorakis, and I. Psaras, “Store edge networkeddata (SEND): a data and performance driven edge storage framework,”inIEEE INFOCOM 2021 - IEEE Conference on Computer Communi-cations (INFOCOM 2021), Vancouver, Canada, May 2021 [resources to be added]
We encourage ideas and developments based on this design and
Otherwise, please browse the source code. It is very well documented and easy to understand.