Agile Objects: Middleware for Survivable Information Systems -- 1999 DARPA ITO Summary

PROJECT SUMMARY

DARPA Order Number E524-1

Contractor:

The Board of Trustees of the University of Illinois 801 South Wright Street Urbana, Illinois 61801

PRINCIPAL INVESTIGATOR

     Andrew A. Chien
     Department of Computer Science
     University of Illinois
     1304 West Springfield Avenue
     Urbana, Illinois 61801
     Phone: 858-822-2458
     Fax:  858-822-2459
     Email: achien@cs.ucsd.edu

Level of Participation 	25%

Related Information

http://www-csag.ucsd.edu/projects/agileO.html

Objective

Agile Objects can increase the flexibility and survivability of high performance distributed systems. The Agile Objects project's goal is to develop a range of technologies which enhance the capabilities of applications based on distributed or component object models transparently. Specifically, Agile Objects will allow component-based applications to be 1) distributed without concern for performance, 2) dynamically redistributed in response to environmental changes, 3) achieve that redistribution while providing hard real-time performance guarantees, and 4) have dynamically changing internal interfaces which increase the difficulty of electronic attack.

Approach

Increasing large-scale use of component object frameworks presents an opportunity for middleware infrastructures which can automatically provide dramatically greater software system flexibility and thereby survivability. We will develop a framework called Agile Objects which leverages component object models and enables the construction of survivable systems that support flexible, online reconfigurability without additional effort by system designers or implementors. The project will produce sound theoretical bases for flexible distributed systems with hard real-time performance guarantees, and significant dynamic reconfiguration capability. For these elements, survivability is enhanced by reconfigurability or "physical agility". In addition, we will explore frameworks for coordinated signature mutation, providing "logical agility" at interfaces, and thereby increasing the difficulty of compromising an interface by removing the binding site for attacking viruses.

To support flexible distributed systems, we will develop high performance distributed object systems which reduce the remote/local invocation performance differences dramatically, allowing object distributions to be performance independent. This not only allows applications to be designed with no specific distributed configuration in mind, but also increases the configuration flexibility viable for hard real-time systems. To support such flexibility and agility while preserving hard real-time performance guarantees, we will extend the open real-time systems theory developed in the Illinois EPIQ project to allow online migration and reconfiguration of application tasks. Finally, we will explore coordinated signature mutation and object migration schemes which increase information survivability without explicit programmer efforts. This approach can be though of as dynamic and automatic wrappers, and enables interfaces to be elusive both in structure and location.

1999 Accomplishments

• Designed, implemented, and demonstrated a series of high performance remote distributed object invocations with performance approaching local RPC invocations. We implemented fast versions of Microsoft's RPC which delivered 105 microsecond and 60 microsecond null RPC round trip times. These represent three to seven-fold improvements in performance, and matches a local invocation using named pipes and approaches that of local invocations using LRPC. Our high performance RPC implementations exploit user level networking, and high speed networks
• Design and construction of a mockup Agile Objects systems for an Aegis battle cruiser which illustrates the potential capabilities of the system and highlights the key challenges in naming, security, resource management, etc. to achieve inherently survivable Agile Object systems. This infrastructure will serve as a driving example in which new functionalities are progressively incorporated as the project progresses.
• Exploration and evaluation of a number of real-time frameworks for Windows NT and Java. Most are lacking in either Win32 or hard real time support. We have not yet selected a final platform for the project due to product uncertainties in this space, and possible systems coming available in the real-time systems research community.

FY 2000 Plans

The major objectives for the fiscal 2000 year for the Agile Objects project are (note this effort ends in 1Q2000):

• Identify/Define a framework for Agile Open Real-time systems and an appropriate real-time infrastructure for developing prototypes of these ideas. Scope includes both commercial and research systems which support "open, composable" real time performance
• Define interfaces and frameworks for key naming, resource management, and security services for Agile Objects based on the driving application mockup of an Aegis battle cruiser. These interfaces and design will be carried forward to succeeding projects on Agile Objects technology.

Technology Transition

• Our high speed RPC systems have been disseminated to and examined by commercial design teams at Sun Microsystems (Javasoft) and Microsoft Research.

Prepared July 1999

Back to CSAG home page