Maruti 3.1 Release Announcement

Maruti is a time-based operating system research project at the University of Maryland. For the past several years we have implemented prototypes of time-based systems as vehicles for further research in development and scheduling of hard real-time systems. In addition to our in-house investigations, Maruti has been used successfully by several research partners.

Now, with Maruti 3.1, we are entering a new phase of our project. We have an operating system suitable for field use by a wider range of users, and we are embarking on the integration of our time-based, hard real-time technology with industry standards and more traditional event-based soft- and non-real-time systems. For this, we are greatly interested in the feedback from users as to the direction of evolution of the system.

For the Maruti 3 project, we will be pursuing the integration of a POSIX interface for soft and non-real-time applications, the use of Ada for Maruti programming, support for asynchronous events and soft/non-real time schedulers within the time-based framework, and heterogeneous Maruti networks.

For this user-oriented phase of the project we will be making regular releases of our software available to allow interested parties to track and influence our development. To begin this phase we are making our current base hard real-time operating system and its development environment available. This is an initial test release.

Runtime System

• The core Maruti runtime system is small - 14 KB code for the single node core, 25 KB code for the distributed core.
• The core provides a calendar-based scheduler, threads, distributed message passing using Time Division Multiplexed Access (TDMA) over the network, and tight time synchronization between network nodes.
• Also included in the runtime system is a graphics library suitable for system monitoring displays as well as simulations.
• Maruti runs on PC-AT compatible computers using the Intel i386 (with i387 coprocessor), i486DX, or Pentium processors. Distributed operation currently requires a 3com 3c507 ethernet card. The graphics library supports standard VGA and Tseng-Labs ET-4000-based Super-VGA. Support for other SVGA chipsets is forthcoming soon.

Development Environment

• The Maruti/Virtual debugging environment - simulates the Maruti runtime system within the development environment. The system clock in this environment tracks virtual time, which can be sped up, slowed down in relation to the actual time, or single-stepped or stopped. This allows temporal debugging of the application. Within Maruti/Virtual traces of the application scheduling and network traffic can be monitored in the debugging session.
• The ANSI-C based Maruti Programming Language (MPL/C). MPL adds modules, message passing primitives, shared memory, periodic functions, message-invoked functions, and exclusion regions to ANSI C. MPL is processed by a version of the GNU C compiler which has been modified to recognize the new MPL features, and to output information about the resources used by the MPL program.
• The Maruti Configuration Language (MCL). MCL allows the system designer to specify the placement, timing constraints, and interconnections of all the modules in an application. MCL is a powerful interpreted C-like language, allowing complex, hierarchical configuration specifications, including replication of components and installation-site specific sizing of the application. The MCL processor analyses the application graph for completeness, and type-checks all connections.
• The Maruti Allocator/Scheduler. The Maruti allocation and scheduling tool analyses the information generated by the MPL compiler and the MCL integrator to find an allocation and scheduling of the tasks of a distributed application across the nodes of a Maruti network. All relative and global timing, exclusion, and precedence constraints are taken into account in finding a schedule, as are the network speed and scheduling parameters.
• The Maruti Timing Trace Analyzer. The Timing Analyzer calculates worst-case computation times from timing files output by the runtime system. Computation times are calculated for each scheduling unit in the application, and these times can be fed back into the Allocator/Scheduler for more precise scheduling analysis.
• The Maruti Runtime Binder (mbind). One of the features of Maruti is the late binding of an application to a particular runtime system. The same application binaries can be combined with different system libraries to build a binary customized for a particular application in a particular setting. Only those portions of the system library needed for that binding are included. Mbind manages this final step.
• The Maruti Application Builder (mbuild). Mbuild automates the process of building an application by generating for the programmer a customizable makefile that manages the complete process of compiling, configuring, scheduling, and binding an application.

Web Accessibility