Overview
VeloxVM is a virtual machine for resource-constrained devices in the Internet of Things. Its primary purpose is to provide a safe and versatile execution environment for small applications on such devices. To this end, VeloxVM features high-level programming languages, preemptive multithreading, exception handling, resource provisioning, and security policies. The services that many IoT operating systems are unable to give to applications because of a lack of hardware and language support can thus be provided by VeloxVM instead. Despite providing a comprehensive feature set, VeloxVM is able to execute together with a host operating system on resource-constrained IoT devices with as little as 32 kB RAM and 256 kB ROM.
Application Development
Applications can be written in the Scheme programming language or Cyclus, which is a new language for IoT application scripting developed in conjunction with VeloxVM. Most of the Scheme standard version R5RS is supported, along with some extensions. Additionally, a considerable number of procedures have been added for network programming and accessing the typical services of an IoT operating system. Cyclus is basically an imperative script language with an infix syntax. When compiling Cyclus scripts, they get translated to Scheme as an intermediary step.
The applications are compiled to a custom bytecode format with high-level instructions. The instruction set contains 191 instructions, and is designed to make IoT apps efficient to express in bytecode. A compact bytecode makes software updates fast and energy-efficient to send over radio, and to store on the devices---either in RAM or in ROM.
Supported operating systems
- POSIX systems such as Linux, *BSD, macOS, and Windows with Cygwin (ports/posix)
- Contiki OS (ports/contiki)
- Contiki-NG (ports/contiki-ng)
Docker setup
The easiest way to get started with VeloxVM is to use a ready-made Docker image available from Docker Cloud. This image contains all of the required software, a compiled VM binary, and compiled example scripts.
For instance, to try the prime number sieve example, run Docker as follows:
docker run nvt1/veloxvm:latest bin/vm apps/sieve.vm
This command will fetch a Docker image that has been generated from the latest pull request in the nvt/veloxvm repository at GitHub. The image will then be cached locally for subsequent runs.
Another possibility is to go into a bash shell, and navigate directly within the Docker container:
docker run -it nvt1/veloxvm:latest bash
Manual setup
Before using the VM for the first time, please ensure that you have the following software installed:
- bison
- flex
- clang
- clisp (or one of the other LISP distributions listed below)
To be able to run all the tests, the following software is also needed:
- afl-fuzz
- python3
- python3-pexpect
The following LISP distributions are supported:
- Armed Bear Common LISP (abcl),
- Clozure CL (ccl)
- CMU Common LISP (cmucl)
- GNU CLISP (clisp)
- Steel Bank Common LISP (sbcl)
By default, the system uses GNU CLISP, but this can be changed in compiler/run.sh by setting the CL_IMPL parameter to either of the values enclosed within parentheses above.
In addition, native compiler executables are supported. To use such an executable, the CL_IMPL value should be set to "native". At the moment, there is only support for creating a native compiler using SBCL. It should also be possible, however, to create a native compiler manually using another LISP implementation and naming the resulting file "compiler".
To create a native compiler based on SBCL, go to compiler/ and run "sbcl --load sbcl-compile.lisp".
Trying out VeloxVM in a POSIX environment
- Build the VM and all tools.
make
- Build the VM apps.
./compile.sh [<app name>]
If no parameter is supplied to the compile.sh script, all apps will
be compiled. The parameter "app name" refers to the name of a app
in the "apps" directory. The name of the app should be without the
suffix in the filename; e.g., ./compile.sh math to
compile a file named either math.scm or math.cyl in the apps
directory.
- Run an app.
bin/vm <app path>
E.g., bin/vm apps/math.vm
Trying out VeloxVM in Contiki OS and Contiki-NG
- Enter the port directory.
For Contiki OS: cd ports/contiki
For Contiki-NG: cd ports/contiki-ng
- Generate a RAM image of the bytecode for an app, in this case "math".
./create-vm-in-ram.sh math
- Build a Contiki firmware with Velox included.
This step requires that we select an appropriate OS platform to build the VM for. We will use the Zoul platform, because it has sufficient RAM and ROM for a basic Velox installation.
make TARGET=zoul vm.upload login
Additional Reading Material
-
N. Tsiftes and T. Voigt. Velox VM: A safe execution environment for resource-constrained IoT applications. Journal of Network and Computer Applications, Volume 118, pages 61-73. 2018.
-
N. Tsiftes. Storage-Centric System Architectures for Networked, Resource-Constrained Devices. Digital Comprehensive Summaries of Uppsala Dissertations from the Faculty of Science and Technology 1331. Uppsala: Acta Universitatis Upsaliensis. 2016.