ChronOS is an
implementation of distributable threads
the Linux kernel, as a first-class abstraction for programming and
scheduling real-time, multi-node sequential behaviors. Distributable
threads first appeared in the Alpha
distributed real-time OS kernel
subsequently in Mach
(a subset), and constitute the first-class programming and scheduling
of the Real-Time
CORBA (Dynamic Scheduling)
ChronOS provides APIs for programming with distributable threads
(similar to that of Real-Time CORBA), algorithms for the real-time
scheduling of distributable threads, a framework for plugging-in
application-specific real-time schedulers for scheduling distributable
threads, and time-critical protocols for
ensuring distributable thread integrity in the presence of failures.
The current version of ChronOS is version 9.06. This is a Beta release.
version scheme is derived from Ubuntu (year.month). ChronOS
on Linux kernel 184.108.40.206 and the real-time linux patch 220.127.116.11.rt27 (
ChronOS currently supports Real-Time CORBA's Case 2 scheduling model
and contains the following algorithms for scheduling distributable
DASA and LBESA are described in Clark's PhD thesis
and Locke's PhD thesis
(CS Department, CMU),
- Fixed priority (RMS)
- Earliest deadline first (EDF)
- Dependent activity scheduling algorithm (DASA)
- Locke's best-effort scheduling algorithm (LBESA)
ChronOS contains the following distributable thread integrity
TPR and D-TPR protocols are described in Curley's MS thesis
(ECE Department, Virginia Tech).
- Thread Polling with Bounded Recovery (TPR)
- Decentralized Thread Polling with Bounded Recovery
ChronOS is available under GNU General
Public License (GPL version 2). The
current release (version 9.06) is available as a patch to Linux kernel
18.104.22.168. ChronOS consists of the following:
- Kernel patch that includes our modifications to the
linux kernel to include distributable thread implementation at the
managment components, and real-time scheduling.
- A middleware service that enables
communication between different nodes for intra-distributable thread
- A middleware library
that provides user-level APIs for programming distributable threads.
The ChronOS version 9.06 kernel patch can be downloaded from here.
The ChronOS version 9.06 middleware can be downloaded from here.
ChronOS version 9.06 is based on kernel 22.214.171.124 and real-time kernel
Download the kernel into a folder, let's call it
- Click here
to download the 126.96.36.199 kernel from kernel.org
- Click here to download the 188.8.131.52
real-time patch from kernel.org
Extract the kernel.
$ tar xjvf linux-184.108.40.206.tar.bz2
Rename the kernel to linux-220.127.116.11-rt27
CONFIG_PREEMPT_RT patch and move
it inside the
kernel directory, and patch the kernel.
# Move the patch inside the kernel folder.
./patch-18.104.22.168-rt27.bz2 | patch -p1
Download the ChronOS ver 9.06 kernel patch as described in the download section put it inside
$ tar zxvf
Compiling the kernel
Install the build essentials.
sudo apt-get install build-essential libncurses5
The kernel should compile fine with
If you get any errors with
that, you can install
create a symbolic link
to it. [optional]
apt-get install gcc-3.4
$ sudo ln -s /usr/bin/gcc-3.4 /usr/bin/gcc
Rename the kernel to chronos-9.06
mv linux-22.214.171.124-rt27 chronos-9.06
Go to the patched ChronOS kernel:
The kernel configuration menu would now come up:
Exit the menu and save the configuration file
- In the "Processor
type and features menu," make sure
Multi-processing support" is unchecked. (ChronOS currently
does not support multiprocessors; multiprocessor support will be
included in a future release.)
- In the "Preemption
Mode" menu, make sure that it is set
to "Complete preemption
If you have more than one core, make sure that you use the
instead. You can:
Once the kernel is compiled, install the modules into /lib/modules:
sudo make modules_install
Installing the kernel
Once the kernel is compiled, install the kernel:
$ sudo cp arch/i386/boot/bzImage
$ sudo cp System.map /boot/System.map-126.96.36.199-
$ sudo cp .config /boot/config-188.8.131.52-
$ sudo update-initramfs -c -k 184.108.40.206-
$ sudo update-grub
ChronOS version 9.06 is installed.
Middleware and library
Download the ChronOS middleware source code package. Create a folder.
Let's call it
middleware service and the library:
$ tar xzvf
$ cd chronos-9.06-middleware/middleware
$ sudo make install
The libraries and the headers are installed in
while the middleware engine
To clean the middleware use:
To un-install the middleware use:
sudo make uninstall
By uninstalling the middleware all the contents of
Installing the remote services
The documentation section describes in detail about creating a remote
service and the configuration file. A sample service is provided inside
the middleware package that was downloaded in the last step.
$ sudo make install
depends on the
and would fail if the
configuration file is not present.
dtengine uses the underlying
network connection. However, on machines
which have more than one interface (
dtengine must be told
which interfaces are being used for distributable threads. ChronOS
would then use those interfaces to communicate with other nodes.
dtengine, make sure that the
following exports are also
$ export LD_LIBRARY_PATH=/usr/local/chronos/lib
# set the network interface (eth0 or eth1)
$ export DT_ETH_PORT=<interface>
These can be added to the
.bashrc to avoid
setting these every time:
$ vi .bashrc
# Add the exports to the end of the .bashrc file
Start the middleware engine:
confirm dtengine is on the path
$ which dtengine
# start the dtengine
- In ChronOS 9.06, the middleware has not been patched
services. Hence, the
middleware needs to be started manually
after every reboot. This would be made into a service in future
- In the current version of ChronOS, only
eth1 are supported. A
more generic version would be available in future releases.
ChronOS architectural stack
The middleware consists of the
and a library
acts like a
background service which
takes care of remote calls from other nodes for invoking services
through the portable interceptor (PI).
order to use distributable threads, you need to
install ChronOS on a minimum of two nodes. Each node would host a set
of "services" that can be called by a distributable thread starting
from one node. A service is essentially a method (i.e., a procedure or
a function) that can be remotely invoked as part of a distributable
Creating (remote) services
The set of services (or methods) is kept in a library file that other
(i.e., machines running ChronOS) can remotely invoke as part of
An example of a service library file looks like as follows:
int node1_service1 (dt_handle_t* handle, void *args)
/* service code */
int node1_service2 (dt_handle_t* handle, void *args)
/* service code */
The service library file must be kept
The services need to be described in a configuration
dtservice.conf file has the required
information about the service:
is a config
FUNC node1_service1 node1_service2
The configuration file has the following details:
name of the service
that has the services or methods
being hosted on the node.
TSER, the total number of methods in
FUNC, the name of the methods
ENDS, the end of the section.
Currently, in ChronOS version 9.06, the name and path of the configuration
file is hard-coded and can't be changed and should be
ChronOS runs a dt-nameserver inside the kernel which receives
information about the location of the services from other nodes.
When the middleware is started on each node, it broadcasts the services
to all other nodes. In the current beta version of ChronOS, the
middleware needs to be restarted every time the services/methods are
changed or updated so that the information is broadcast to every
A sample service along with a Makefile is available inside the
APIs and System calls
libdt.so provides the APIs that can be
used for programming with distributable threads:
(dt_handle_t *handle, void (*function)(dt_handle_t*, void*), void
int dt_remote_invoke (dt_handle_t*, const char *name, void *args);
The header file
function prototypes and data structures.
Documentation about the middleware and how to create a test application
can be downloaded here.
dt_cancel, dt_pause, and real-time
scheduling functions are available but these are not fully
the current beta release.
These would be available in the future release.
are available here: user's
guide and programmer's
Documentation about real-time scheduling algorithms
implemented in ChronOS and how they can be extended can be downloaded here.
Documentation about thread integrity protocols
implemented in ChronOS are available here: user's
guide and overview.
Distributable thread middleware
Distributable thread integrity protocols
- Recovering from Distributable Thread Failures in Distributed Real-Time Java, E. Curley, B. Ravindran, J. Anderson, and E. D. Jensen, ACM Transactions on Embedded Computing Systems (TECS), To appear
- Recovering from Distributable Thread
Assured Timeliness in Real-Time Distributed Systems, Edward
Master's Thesis, ECE Department,
Virginia Tech, Blacksburg, VA, USA, February 2007
- On Best-Effort Real-Time Assurances for
Recovering from Distributable Thread Failures, B. Ravindran,
E. Curley, J. Anderson, and E. D. Jensen, IEEE International Symposium on
Object and component-oriented Real-time distributed Computing (ISORC),
pages 344-353, May 2007
Integrity Protocols for Distributable Real-Time Threads in Dynamic
Distributed Systems, B. Ravindran, E. Curley, J. Anderson,
and E. D. Jensen, International
Workshop on Embedded Software Optimization (ESO), IFIP International
Conference on Embedded and Ubiquitous Computing (EUC),
pages 660-673, LNCS, Volume 4809/2007,
Springer, December 2007
- Recovering from Distributable Thread
Failures with Assured Timeliness in Real-Time Distributed Systems,
E. Curley, J. Anderson, B. Ravindran, and E. D. Jensen, IEEE Symposium on Reliable
Distributed Systems (SRDS), pages: 267 - 276, October 2006
Distributable thread scheduling
Networked Embedded Systems, J. Anderson, B. Ravindran, and E.
Jensen, IFIP International Conference on Embedded and
Computing (EUC), pages 247-260, LNCS, Volume 4808/2007,
Springer, December 2007
Distributable Real-Time Threads
Networked Embedded Systems in the Presence of Crash Failures and
Message Losses, S. Fahmy, B. Ravindran, and E. D. Jensen, ACM
On Applied Computing (SAC), Track on Real-Time Systems,
pages 294-301, March 2008
- On collaborative scheduling of distributable
real-time threads in dynamic, networked embedded systems, S.
Ravindran, and E. D. Jensen, IEEE International Symposium on
component-oriented Real-time distributed Computing (ISORC),
pages 485-491, May 2008
Scheduling of Distributable Real-Time
with Assured End-to-End Timeliness, S. Fahmy, B. Ravindran,
and E. D.
Jensen, 13th International Conference on Reliable Software
- Ada-Europe, pages 211-225, LNCS, Volume 5026, Springer,
June 2008 (full version of the paper is available
as a technical report)
- Scheduling Dependent Distributable Real-Time
in Dynamic Networked Embedded Systems, S. Fahmy, B.
Ravindran, and E.
D. Jensen, IFIP Working Conference on Distributed and
Systems (DIPES), pages 171-180, Distributed
Embedded Systems: Design, Middleware, and Resources, B.
Kleinjohann, L. Kleinjohann, and W. Wolf (editors), Springer, Volume
271, September 2008 (full version of the paper is available
as a technical report)
- On Distributed Real-Time Scheduling in
Networked Embedded Systems in the Presence of Crash Failures,
B. Ravindran, J. Anderson, and E. D. Jensen, IFIFP Workshop on Software
Technologies for Future Embedded and Ubiquitous Systems (SEUS), IEEE
ISORC, May 2007
- Probabilistic, Real-Time Scheduling of
Distributable Threads Under Dependencies in Ad Hoc Networks,
K. Han, B. Ravindran, and E. D. Jensen, IEEE Wireless Communications and
Networking Conference (IEEE WCNC), pages 2253-2258, March
- Exploiting Slack for Scheduling Dependent,
Distributable Real-Time Threads in Mobile Ad Hoc Networks, K.
Han, B. Ravindran, and E. D. Jensen, International Conference on
Real-Time and Network Systems (RTNS), pages 225-234, March
- RTG-L: Dependably Scheduling Real-Time
Distributable Threads in Large-Scale, Unreliable Networks, K.
Han, B. Ravindran, and E. D. Jensen, IEEE Pacific Rim International
Symposium on Dependable Computing (PRDC), pages 314-321,
The ChronOS project is led by Prof. Binoy Ravindran at Virginia
Students of Virginia Tech's Real-Time Systems
Laboratory are leading
different aspects of the implementation effort:
Contact Piyush Garyali (firstname.lastname@example.org)
or Prof. Ravindran (email@example.com)
for installation problems, questions or comments.
- Matthew Dellinger
- Piyush Garyali
- Caleb Hopkins
- Surjyadipta Shome