OptIPuter Software: System Software for High Bandwidth Network Applications

 

Motivation  The advent of low-cost, plentiful wide area bandwidth based on dense wavelength division multiplexing (DWDM) technology coupled with plentiful storage and computing presents new opportunities for applications and new challenges in systems design.  The progress of Moore’s Law over the past decade and its impact on computing has produced a revolution in how we use and deploy computing – ranging across servers, desktops, laptops, handhelds, and increasing numbers of embedded devices (sensors, phones, etc).  Remarkably, storage and network technologies are increasing in capability far faster than Moore’s law, feeding a revolution in potential capability of comparable scale.  The bandwidth and storage exponentials crossing Moore™s Law fundamentally alters a wide range of system design tradeoffs – that which was scarce is now abundant and vice versa. The implications of this change for systems design are legion.  The scope for the OptIPuter project is to explore these changes to pioneer innovative system software services and applications.

 

Objectives To explore how this changing technological capability enables exciting new applications and drives a rethinking of systems design, we are researching:

 

-         Distributed services, data abstractions, data management, and APIs for a bandwidth rich environment

-         Networking protocols, implementations, and presentation layers for a Terabit (100 x 10Gb) environment

-         Parallel communication software layers, operating at maximum system hardware speeds

-         Security models and protocols for DWDM-based distributed systems

-         Switch configuration and management algorithms, policies, and mechanisms

-         Demonstrations of these software systems on large-scale scientific applications

 

The larger OptIPuter project includes a number of activities which support our research goals:

-         the construction of a wide area testbed of DWDM-based networks with the capability to configure lambdas in real-time, scalable Linux clusters, and large data resources

-         distributed visualization and collaboration applications for a bandwidth-rich environment

-         using two leading edge applications – Neurosciences Data Analysis and Earth Sciences Data Analysis – both of which require online visualization of terabytes of data to drive designs and demonstrations

 

Project Members:  Xinran (Ryan) Wu, Huaxia Xia, Justin Burke, Nut Taesombut, Eric Weigle, and Andrew Chien

 

Objectives and Activities

-         Programming Model / Resource Models for OptIPuter systems

1.      Functional, performance, and security abstractions for distributed collections of resources

2.      “Distributed Virtual Computers”, trusted and untrusted sets of resources

3.      Implementation technologies for these Resource Models

-         High Speed Group Communication (Group Transport Protocol – GTP) for OptIPuter systems

1.      evaluating high speed data movers (GridFTP, etc.)

2.      evaluating high speed protocols over IP (XCP, etc.)

3.      Design of new network transport protocols and experiments (GTP)

-         High Speed Robust Storage for OptIPuter

1.      Exploring massively parallel access to distributed storage

2.      Protection and security models

3.      Approaches for robust sharing and performance

-         Software Architecture for OptIPuter (see papers)

 

Papers

If you are a UCSD graduate student, and interested in joining the OptIPuter project, email Andrew Chien.  More information on OptIPuter can be found on the overall project page.  Other institutions involved in this project include UCSD CalIT2, UI Chicago, USC Information Sciences Institute, UC Irvine, Northwestern Univ, San Diego State, Scripps Institute of Oceanography, and SAIC.


Funding Information

The OptIPuter effort  is supported by the National Science Foundation as part of the Information Technology Research (ITR) program under NSF Cooperative Agreement ANI-0225642.

Last updated, May 2004
 

Back to CSAG home page

webmaster