2000 Spring SIW
       Abstract Submittal Form


Primary Author Information:  Harry Wolfson

MIT Lincoln Laboratory
244 Wood St.
Lexington, MA 02421  USA
781-981-5996
HarryWolfson@LL.MIT.EDU


Other Authors: none


PROPOSED TITLE:  "Flow Control in Reliable Multicast"


Keywords: RTI, Reliable, Multicast, flow control, squelch, throughput

Forum you are targeting:   SIW-INF Track: SIW-RTI Forum 


ABSTRACT:

The High Level Architecture (HLA) Runtime Infrastructure (RTI) provides
a variety of message Transport Services for different communication
needs. These Transport Services include Best Effort and Reliable. While
Best Effort Multicast messaging is fairly well standardized using UDP
(User Datagram Protocol), a protocol for Reliable Multicast is still a
technology in development.

The Reliable Multicast service that was first implemented for the STOW
Program has had significant development over the years to enhance its
utility for a wide variety of users, including improved robustness and
reliability for delivery of messages. The most recent version of this
Reliable message service is currently being used in the DMSO RTI
Prototype.

This paper will describe the recent addition of Flow Control to the RTI
Reliable service. End to end flow control prevents rapid, or heavy,
producers of data from overwhelming and clogging the network. It insures
that subscribers are able to receive and process that data in a timely
manner without overflowing their own buffers, or preventing other
traffic on the network from continuing to flow.



Summary of the major points you intend to convey:

Prior implementations of Reliable Multicast handled extremely fast/heavy
producers of data, combined with slower consumers of data, in a less
than ideal manner. Cached data would fill available buffers; or
socket-based connections would stall and/or die; or data would simply be
dropped. This paper will describe a new protocol that allows all
producers and consumers of data in a Federation to coordinate the flow
of data sent via Reliable Multicast service. This flow control: helps
prevent clogged networks (deadlock, deadly embrace, etc.); prevents the
uncontrolled filling of buffers and available memory; prevents the
discarding of data due to the network, or a single Federate, being
overwhelmed with too much data. In other words, Flow Control for
Reliable Multicast in the DMSO RTI Prototype provides a control on the
throughput of data to match the capabilities of the various Federates
within an Execution.