-
J. Field and C. A. Varela.
Toward a Programming Model for Building Reliable Systems with Distributed State.
Electronic Notes in Theoretical Computer Science,
68(3):1-19,
March 2003.
Note: Invited journal paper.
Keyword(s): distributed computing,
concurrent programming,
coordination models.
Abstract: |
We present the preliminary design of a programming model for building reliable systems with distributed state from collections of potentially unreliable components. Our transactor model provides constructs for maintaining consistency among the states of distributed components. Our intention is that transactors should support key aspects of both traditional distributed transactions, e.g., for electronic commerce, and systems with weaker consistency requirements, e.g., peer-to-peer file- and process-sharing systems. In this paper, we motivate the need for language support for maintenance of distributed state, describe the design goals for the transactor model, provide an operational semantics for a simple transactor calculus, and provide several examples of applications of the transactor model in a higher-level language. |
@Article{field-varela-transactors-entcs-2003,
author = {J. Field and C. A. Varela},
editor = {A. Brogi and J.M. Jacquet},
title = {Toward a Programming Model for Building Reliable Systems with Distributed State},
journal = {Electronic Notes in Theoretical Computer Science},
year = 2003,
publisher = {Elsevier},
volume = 68,
number = 3,
pages = {1--19},
month = {March},
note = {invited journal paper},
url = {http://www.elsevier.nl/locate/entcs/volume68.html},
pdf = {http://www.cs.rpi.edu/~cvarela/entcs2003.pdf},
keywords = {distributed computing, concurrent programming, coordination models},
abstract = {We present the preliminary design of a programming model for building reliable systems with distributed state from collections of potentially unreliable components. Our transactor model provides constructs for maintaining consistency among the states of distributed components. Our intention is that transactors should support key aspects of both traditional distributed transactions, e.g., for electronic commerce, and systems with weaker consistency requirements, e.g., peer-to-peer file- and process-sharing systems. In this paper, we motivate the need for language support for maintenance of distributed state, describe the design goals for the transactor model, provide an operational semantics for a simple transactor calculus, and provide several examples of applications of the transactor model in a higher-level language.}
}
-
Kaoutar El Maghraoui,
J. Flaherty,
B. Szymanski,
J. Teresco,
and Carlos A. Varela.
Adaptive Computation over Dynamic and Heterogeneous Networks.
In Proc. of the Fifth International Conference on Parallel Processing and Applied Mathematics (PPAM'2003),
number 3019 of LNCS,
Czestochowa, Poland,
pages 1083-1090,
September 2003.
Keyword(s): grid computing,
middleware.
Abstract: |
Over the last two decades, efficient message passing libraries have been developed for parallel scientific computation. Concurrently, programming languages have been created supporting dynamically reconfigurable distributed systems over the heterogeneous Internet. In this paper, we introduce SALSA-MPI, an actor programming language approach to scientific computing that extends MPI with a checkpointing and migration API and a runtime system that manages both periodic checkpoints and process or application migration. The goal is to enable dynamic network reconfiguration and load balancing without sacrificing application performance or requiring extensive code modifications. As driving technology for this effort of unifying parallel and distributed computing, we plan to use adaptive solvers of partial differential equations. Fields as diverse as fluid dynamics, material science, biomechanics, and ecology make use of parallel adaptive computation, but target architectures have traditionally been supercomputers and tightly-coupled clusters. SALSA-MPI is intended to allow these computations to make efficient use of more distributed and dynamic computing resources. |
@InProceedings{elmaghraoui-salsampi-ppam-2003,
author = {Kaoutar El Maghraoui and J. Flaherty and B. Szymanski and J. Teresco and Carlos A. Varela},
title = {Adaptive Computation over Dynamic and Heterogeneous Networks},
booktitle = {Proc. of the Fifth International Conference on Parallel Processing and Applied Mathematics (PPAM'2003)},
year = 2003,
number = 3019,
pages = "1083--1090",
series = {LNCS},
address = {Czestochowa, Poland},
month = {September},
pdf = {http://wcl.cs.rpi.edu/papers/elmagk03.pdf},
keywords = {grid computing, middleware},
abstract = {Over the last two decades, efficient message passing libraries have been developed for parallel scientific computation. Concurrently, programming languages have been created supporting dynamically reconfigurable distributed systems over the heterogeneous Internet. In this paper, we introduce SALSA-MPI, an actor programming language approach to scientific computing that extends MPI with a checkpointing and migration API and a runtime system that manages both periodic checkpoints and process or application migration. The goal is to enable dynamic network reconfiguration and load balancing without sacrificing application performance or requiring extensive code modifications. As driving technology for this effort of unifying parallel and distributed computing, we plan to use adaptive solvers of partial differential equations. Fields as diverse as fluid dynamics, material science, biomechanics, and ecology make use of parallel adaptive computation, but target architectures have traditionally been supercomputers and tightly-coupled clusters. SALSA-MPI is intended to allow these computations to make efficient use of more distributed and dynamic computing resources.}
}
-
B. Szymanski,
Carlos A. Varela,
J. Cummings,
and J. Napolitano.
Dynamically Reconfigurable Scientific Computing on Large-Scale Heterogeneous Grids.
In Proc. of the Fifth International Conference on Parallel Processing and Applied Mathematics (PPAM'2003),
volume 3019 of LNCS,
Czestochowa, Poland,
pages 419-430,
September 2003.
Note: Invited paper.
Keyword(s): distributed computing,
grid computing.
Abstract: |
Many scientific applications require computational capabilities not easily supported by current computing environments. We propose a scalable computing environment based on autonomous actors. In this approach, a wide range of computational resources, ranging from clusters to desktops and laptops, can run an application programmed using actors as program components in an actor language: SALSA. SALSA actors have the ability to execute autonomously in dynamically reconfigurable computing environments. We develop the corresponding “Internet Operating system” (IO) to address run-time middleware issues such as permanent storage for results produced by actors, inter-actor communication and synchronization, and fault-tolerance in a manner transparent to the end-user. We are using this worldwide computing software infrastructure to solve a long outstanding problem in particle physics: the missing baryons, originally identified over thirty years ago. |
@InProceedings{szymanski-iosphysics-ppam-2003,
author = {B. Szymanski and Carlos A. Varela and J. Cummings and J. Napolitano},
title = {Dynamically Reconfigurable Scientific Computing on Large-Scale Heterogeneous Grids},
booktitle = {Proc. of the Fifth International Conference on Parallel Processing and Applied Mathematics (PPAM'2003)},
year = 2003,
number = 3019,
pages = {419-430},
volume = {3019},
series = {LNCS},
address = {Czestochowa, Poland},
month = {September},
note = {invited paper},
ps = {http://wcl.cs.rpi.edu/papers/IPszymanski.ps},
keywords = {distributed computing, grid computing},
abstract = {Many scientific applications require computational capabilities not easily supported by current computing environments. We propose a scalable computing environment based on autonomous actors. In this approach, a wide range of computational resources, ranging from clusters to desktops and laptops, can run an application programmed using actors as program components in an actor language: SALSA. SALSA actors have the ability to execute autonomously in dynamically reconfigurable computing environments. We develop the corresponding “Internet Operating system” (IO) to address run-time middleware issues such as permanent storage for results produced by actors, inter-actor communication and synchronization, and fault-tolerance in a manner transparent to the end-user. We are using this worldwide computing software infrastructure to solve a long outstanding problem in particle physics: the missing baryons, originally identified over thirty years ago.}
}
-
Robin Toll and Carlos A. Varela.
Mobility and Security in Worldwide Computing.
In Proceedings of the 9th ECOOP Workshop on Mobile Object Systems,
Darmstadt, Germany,
July 2003.
Keyword(s): distributed computing,
grid computing,
software agents,
internet programming languages.
Abstract: |
Modern distributed computing requires a secure framework capable of free code mobility. In this paper, we present a simple lambda-based actor language with extensions for mobility and security, as well as the operational semantics to reason about these topics in distributed systems. Finally, we describe our preliminary implementation results |
@InProceedings{toll-varela-smal-2003,
author = {Robin Toll and Carlos A. Varela},
title = {Mobility and Security in Worldwide Computing},
booktitle = {Proceedings of the 9th ECOOP Workshop on Mobile Object Systems},
year = 2003,
address = {Darmstadt, Germany},
month = {July},
pdf = {http://wcl.cs.rpi.edu/papers/ecoopws2003Full.pdf},
keywords = {distributed computing, grid computing, software agents, internet programming languages},
abstract = {Modern distributed computing requires a secure framework capable of free code mobility. In this paper, we present a simple lambda-based actor language with extensions for mobility and security, as well as the operational semantics to reason about these topics in distributed systems. Finally, we describe our preliminary implementation results}
}
-
Harihar N. Iyer,
Abe Stephens,
Travis Desell,
and Carlos A. Varela.
OverView - Dynamic Visualization of Java-Based Highly Reconfigurable Distributed Systems.
Technical report,
Rensselaer Polytechnic Institute Worldwide Computing Laboratory,
August 2003.
Keyword(s): distributed computing,
distributed systems visualization.
Abstract: |
Online visualization enables developers to test, debug, and monitor the behavior of distributed systems, while they are running. While important in software development, online visualization of distributed systems is largely unaddressed by conventional tools. Distributed systems are often programmed using high-level abstractions that facilitate reasoning about them, e.g., actors, processes, sessions, or ambients. OverView is an entity specification language-driven Eclipse plug-in for visualization of distributed systems that preserves the high level of abstraction, and enables online visualization of critical distributed system properties such as component naming, location, remote communication, and migration. OverView’s architecture is generic in that different abstractions can reuse the visualization module requiring changes only in the entity specifications that drive the visualization process. |
@TechReport{iyer-overview-2003,
author = {Harihar N. Iyer and Abe Stephens and Travis Desell and Carlos A. Varela},
title = {OverView - Dynamic Visualization of Java-Based Highly Reconfigurable Distributed Systems},
institution = {Rensselaer Polytechnic Institute Worldwide Computing Laboratory},
year = 2003,
month = {August},
pdf = {http://wcl.cs.rpi.edu/papers/overview2003.pdf},
keywords = {distributed computing, distributed systems visualization},
abstract = {Online visualization enables developers to test, debug, and monitor the behavior of distributed systems, while they are running. While important in software development, online visualization of distributed systems is largely unaddressed by conventional tools. Distributed systems are often programmed using high-level abstractions that facilitate reasoning about them, e.g., actors, processes, sessions, or ambients. OverView is an entity specification language-driven Eclipse plug-in for visualization of distributed systems that preserves the high level of abstraction, and enables online visualization of critical distributed system properties such as component naming, location, remote communication, and migration. OverView’s architecture is generic in that different abstractions can reuse the visualization module requiring changes only in the entity specifications that drive the visualization process.}
}
-
Camron Tolman.
A Fault-Tolerant Home-Based Naming Service for Mobile Agents.
Master's thesis,
Rensselaer Polytechnic Institute,
April 2003.
Keyword(s): distributed computing,
grid computing,
middleware,
software agents,
concurrent programming.
Abstract: |
As a mobile software agent migrates to various host machines on a network, collaborating agents need to be able to locate the mobile agent for communication. A naming service is in charge of locating such an agent in a distributed system given its name. Three critical characteristics of a naming service are: fault tolerance, scalability, and efficient name resolution. Most naming services provide support for efficient name resolution but do not address fault tolerance or scalability issues. Conversely, distributed hash-table approaches to naming provide fault tolerance and scalability albeit at a sacrificed name resolution performance. We introduce a Fault-Tolerant Home-Based Naming Service (FHNS) that is robust and scalable yet enabling efficient name resolution. An agent using FHNS has a unique name that encodes its home base. Home bases are connected in a peer-to-peer manner. The agent’s run-time system is responsible for keeping the agent’s home base informed of its network location. An agent obtains a reference to another agent by knowing the name of that agent and requesting any home base to resolve that agent’s location. Redundancy exists between neighboring home bases to ensure the location of any given agent can still be resolved despite an agent’s respective home base failing, thus eliminating single points of failure. In the event of a home base failure, the failover procedure occurs transparently to the agents. Furthermore, FHNS has optimal sequential fault tolerance behavior: if the home bases fail sequentially down to any single remaining home base, FHNS is still able to resolve the location of every agent in the distributed system. Resolving the location of an agent with a distributed hash-table approach, such as Chord or SPRR, requires O(log n) messages between home bases. In a worldwide computing system with potentially billions of nodes, such a lookup could still take up to 30 messages. Realizing that logarithmic lookups are not as efficient as needed for a naming service, FHNS provides name to location resolution in one round-trip request (two messages) in the general case. A logarithmic number of messages is only needed when a direct request to a home base fails. FHNS is a novel contribution being made to the field of mobile agent computing. Fault-tolerant home-based naming can be incorporated into existing mobile agent platforms to eliminate single points of failure without sacrificing efficient name resolution. |
@MastersThesis{tolman-fhns-2003,
author = {Camron Tolman},
title = {A Fault-Tolerant Home-Based Naming Service for Mobile Agents},
school = {Rensselaer Polytechnic Institute},
month = {April},
year = 2003,
pdf = {http://wcl.cs.rpi.edu/theses/fhns/cam_thesis_final.pdf},
keywords = {distributed computing, grid computing, middleware, software agents, concurrent programming},
abstract = {As a mobile software agent migrates to various host machines on a network, collaborating agents need to be able to locate the mobile agent for communication. A naming service is in charge of locating such an agent in a distributed system given its name. Three critical characteristics of a naming service are: fault tolerance, scalability, and efficient name resolution. Most naming services provide support for efficient name resolution but do not address fault tolerance or scalability issues. Conversely, distributed hash-table approaches to naming provide fault tolerance and scalability albeit at a sacrificed name resolution performance. We introduce a Fault-Tolerant Home-Based Naming Service (FHNS) that is robust and scalable yet enabling efficient name resolution. An agent using FHNS has a unique name that encodes its home base. Home bases are connected in a peer-to-peer manner. The agent’s run-time system is responsible for keeping the agent’s home base informed of its network location. An agent obtains a reference to another agent by knowing the name of that agent and requesting any home base to resolve that agent’s location. Redundancy exists between neighboring home bases to ensure the location of any given agent can still be resolved despite an agent’s respective home base failing, thus eliminating single points of failure. In the event of a home base failure, the failover procedure occurs transparently to the agents. Furthermore, FHNS has optimal sequential fault tolerance behavior: if the home bases fail sequentially down to any single remaining home base, FHNS is still able to resolve the location of every agent in the distributed system. Resolving the location of an agent with a distributed hash-table approach, such as Chord or SPRR, requires O(log n) messages between home bases. In a worldwide computing system with potentially billions of nodes, such a lookup could still take up to 30 messages. Realizing that logarithmic lookups are not as efficient as needed for a naming service, FHNS provides name to location resolution in one round-trip request (two messages) in the general case. A logarithmic number of messages is only needed when a direct request to a home base fails. FHNS is a novel contribution being made to the field of mobile agent computing. Fault-tolerant home-based naming can be incorporated into existing mobile agent platforms to eliminate single points of failure without sacrificing efficient name resolution.}
}