Browse Prior Art Database

Multicasting Enhancement to Distributed Computer Environment/Remote Processing Call Semantics for Distributed Application Client

IP.com Disclosure Number: IPCOM000117856D
Original Publication Date: 1996-Jun-01
Included in the Prior Art Database: 2005-Mar-31
Document File: 4 page(s) / 151K

Publishing Venue

IBM

Related People

Lin, DH: AUTHOR

Abstract

Current OSF Distributed Computing Environment/Remote Processing Call (DCE/RPC) provides four semantics: idempotent, non-idempotent (at-most-once), maybe, and broadcast for distributed applications. Except broadcast RPC, all other semantics are based on one-to-one mapping, namely one client making one call to one server at one time.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 38% of the total text.

Multicasting Enhancement to Distributed Computer Environment/Remote
Processing Call Semantics for Distributed Application Client

      Current OSF Distributed Computing Environment/Remote Processing
Call (DCE/RPC) provides four semantics:  idempotent, non-idempotent
(at-most-once), maybe, and broadcast for distributed applications.
Except broadcast RPC, all other semantics are based on one-to-one
mapping, namely one client making one call to one server at one time.

The support for one client talking to several servers simultaneously
is very minimum (e.g., using multi-threaded client calls).  Although
broadcast RPC has capability of one-to-n mapping, it doesn't really
know/care who the servers will be and what's the result of the call.
This disclosure proposed a  multicast semantics extension to the
DCE/RPC to provide an ease-of-use parallel processing capability for
distributed application client.

      Distributed computing is popular in modern computing
environment, especially after OSF DCE introduced in the 1990s.  The
basic concept behind the OSF DCE environment is to share resources
among all the machines configured in a network environment.  It is
reasonable to assume that the RPC semantics provided in DCE are
sufficient enough to achieve distributed sharing of the system
resources.  However, the configured network machines sometime can
also be utilized in a parallel fashion where all the server machines
may become part of the processing unit of a single application.  As a
result, distributed application can be executed in parallel to
achieve higher application performance and system throughput.

      In DCE/RPC, the provided semantics - idempotent, non-idempotent
(at-most-once), maybe, and broadcast, do not have direct support to
the extension of parallel execution of distributed client application
(although multi-threaded capability is provided, it is most used in
server application).  This limitation may impact DCE to be exploited
into the parallel computing arena.  This disclosure proposes a
multicasting enhancement to the DCE/RPC semantics in order to widen
the adoption of DCE to the future computing environment.

Multicast RPC Design

      In the current DCE, making each RPC call on the client side
literally need to go through following steps:  find out compatible
server bindings from directory service, fill in the data structure
with requested interface information, set up RPC call handle
(includes marshalling and authentication if needed), fire the RPC
call, and wait for the reply from the server side.  Of course, most
of these actions are taken care of by the client side stub code.
Basically, each RPC call requires three main arguments-server binding
requesting interface, and RPC call handle to complete the call.
Therefore, in order to introduce a new call semantic on the client
side, the three arguments mentioned before required some design work.

      Obviously, the proposed Multicast RPC re...