Features:

HPC INTERVIEW WITH DR. ANDREW CHIEN, UCSD
By Alan Beck, Editor-in-Chief, HPCwire

Q: I understand that you are building a new type of Grid computing environment called OptIPuter and plan to integrate BigBangwidth's LightPath Accelerator technology?

A: The OptIPuter, named for Optical networking, Internet Protocol, computer storage, processing and visualization technologies, is an infrastructure that tightly couples computational resources over parallel optical networks using the IP communication mechanism. Funded by a $13.5 million grant from the National Science Foundation, the project is a collaboration of institutions both in the United States and internationally. The OptIPuter exploits a new world in which the central architectural element is optical networking, not computers -- creating "supernetworks."

BigBangwidth's LightPath Accelerator technology is an all-optical networking solution that can accelerate network performance within any LAN or SAN protocol. We are exploring the implications of such technologies for distributed computational science. The BigBangwidth switch provides interesting opportunities both for the fundamental systems research we are pursuing, as well as enabling new capabilities for the applications.

Q: What is the goal of the OptIPuter project and how does BigBangwidth's technology help your research team?

A: The goal of the OptIPuter project is to enable scientists who are generating terabytes and petabytes of data to interactively visualize, analyze and correlate their data from multiple storage sites connected by optical networks.

BigBangwidth's technology supports important parts of the OptIPuter vision. It can detect large flows automatically or be manually controlled to configure on-demand private lightpaths for an individual user or application. This enables a wealth of storage area network and general networking experiments. Because it is a transparent optical interconnect, there's no incremental cost for moving to higher speeds such as 10 gigabits.

"We are very excited about participating with the University of California San Diego and the OptIPuter project. Enabling researchers to push scientific applications has significant benefits and is also very rewarding" said Dan Gatti, president and CEO of BigBangwidth.

Q: Can you describe to our readers your network architecture including the LightPath Accelerator?

A: We are pursuing two uses for the LightPath Accelerator. Initially, we are studying the integration of SANs and optical switches such as BigBangwidth's. The switch will be used to couple large storage arrays and high speed scientific computing clusters at high speeds, including both 1 Gig and 10 Gig links.

We are also considering novel ways to apply this technology in metro and wide-area networks. One possibility is to use it to complement more expensive routed or packet-switched infrastructures, sharing ports or as a high-speed bypass.

Q: Why did you choose BigBangwidth's LightPath Accelerator?

A: First, LightPath Accelerator works with IP and Ethernet, so we can easily integrate it with our commodity clusters and storage systems. They also have a nice software solution, allowing the circuit switching to be incorporated in a seamless fashion even for systems that expect packet-switched, fixed circuit connection behavior.

Second, the transparent optical circuit switching offers scaling to higher speeds, protocol independence and reach to significant distances without adding expensive amplification. In short, LightPath Accelerator will enable us to couple storage to high-bandwidth visualizations at 10 Gigabits across the campus at a modest cost. A further benefit is that the tremendous bandwidth and dedicated connection avoids lost frames, and late frames, and variability in performance.

Finally, BigBangwidth provides an API that allows us to apply the basic LightPath Accelerator components in novel ways to our own research.

Q: How do you expect these products will impact applications?

A: A core idea in the OptIPuter project is that new levels of dedicated bandwidth enable new levels of capability in Grids, simulation and visualization. For example, the OptIPuter is focused on enabling gigapixel visualizations based on gigazone simulations and large-scale sensor data collection which are now becoming commonplace. In short, the data is available, but the bottleneck is the network. With lightpath networking, it is becoming possible to get the data between the compute engines, and onto the display devices at reasonable cost.

Q: What other institutes are involved in the OptIPuter project? And, how will they each utilize LightPath Accelerator?

A: The OptIPuter project includes researchers from the University of California-San Diego, the University of Illinois-Chicago, the California Institute for Telecommunications and Information Technology, Northwestern University, San Diego State University, the University of Southern California, the University of California-Irvine and Texas A&M University.

We also have affiliates, advisors or collaborators from the University of Amsterdam, the Scripps Institute of Oceanography, the U.S. Geological Survey, the University of California-Santa Barbara, the Monterey Bay Aquarium, NASA Ames, the San Diego Supercomputer Center and the National Partnership for Advanced Computational Infrastructure. Finally, we also have numerous commercial affiliations, including Chiaro Networks, IBM, Sun Microsystems and Telcordia Technologies.

The LightPath Accelerator equipment will become part of our shared infrastructure, and enable a broad range of systems software and applications research within the OptIPuter project. If we are successful in developing innovative middleware which automatically manages the underlying optical fabric, all of the applications and users will benefit from this infrastructure –- with no explicit effort.

Q: Do you anticipate near-term commercial applications resulting from your research such as in the area of NAS and SAN, and if so, can you share your thoughts with our readers?

A: The most direct near-term application is likely to be SAN bridging over the wide area links. With transparent optical switches and private lambdas, many corporations will be able to build distributed virtual machines at reasonable cost. In the longer term, the the OptIPuter vision is much broader, so its hard to anticipate the exciting new applications that might emerge.