Simon Kahan, PhD

Affiliate Faculty

Research Scientist - High Performance Computing, Northwest Institute for Advanced Computing

Dr. Kahan is also an Affiliate Faculty member in Computer Science and Engineering at the University of Washington.

Dr. Kahan has extensive experience with high performance computer (HPC) systems. A first conversation with ISB faculty member Dr. Shmulevich and affiliate Dr. Flann in 2010 began with, “Where do you think HPC could have the greatest impact on your work?”  Their response led Kahan to pursue funding from the Pacific Northwest National Laboratory (PNNL) needed to create Biocellion. Biocellion simulates the behavior of living systems having billions of cells at high fidelity. ISB uses Biocellion to model yeast colonies, human skin, and glioma. In late 2013, Kahan formed Biocellion SPC to further adoption.

For the twenty­five years prior, Kahan had held various technical and management positions at Pacific Northwest National Laboratory, Google, Cray, Tera Computer, and AT&T Bell Laboratories. He was educated as an Electrical Engineer at UC Berkeley and as a Computer Scientist at the University of Washington in the 1980s. Kahan’s contributions center around innovations that overcome the congestion and inefficient use of resources that arise within large computer systems applied to solve big problems. Curious about how these methods compare to those applied in other domains involving systems, he returned to school at Seattle University, where he studied emergence of collective vision through their Organization Systems Renewal program.

At about the same time as starting Biocellion, Kahan started a project at the University of Washington in Computer Science and Engineering (“Grappa”) based on earlier work at Tera and Google. Grappa is a general purpose software platform intended to offer high performance particularly in the space of irregular applications, which includes large graph analytics. Graph analytics centers on discovering patterns of relationship such as arise in systems biology and other domains where knowledge is represented by networks.

Grappa has conceptual similarities to methods for facilitating emergence of collective vision. Both invest time in transformational processes that introduce disorder to create opportunity for better alignment of subsystem behaviors before execution. The end result is greater efficiency within system components.

In social systems, stakeholders with diverse perspectives may intentionally be assembled despite the conflict that ensues. A time-consuming technique called “subgrouping” is used to establish common ground. From this common ground emerges a collective vision that provides the foundation for a sustainable solution. Aligned in their vision, stakeholders’ subsequent actions contribute more directly to the solution.

Grappa programs produce enormous numbers of sub-computations with diverse resource requirements. Grappa expends time to align the needs of these sub-computations. Though this slows down progress of the individual sub-computations, greater computational efficiency results from the communication, processing, and memory subsystem components. A consequence is that for several data-intensive applications and scientific benchmarks, the University of Washington team has shown that Grappa outperforms contemporary approaches even on relatively small computer clusters having only a few hundred processors.

Social systems and HPC are not the only domains where creating opportunity through disorder at the system level creates opportunities to realign and achieve greater efficiency at the subsystem or component level. Additional examples are numerical optimization and evolutionary biology. In a pre-existing system, the purpose, system, subsystems and their desired order are stipulated: at most the inducement of disorder and, to a greater extent, the design of the transformative processes are controllable interventions. Kahan shares other engineers’ and scientists’ interests is designing such interventions. Prerequisite to that design is understanding not only the subsystems’ properties but also their relationships within the encompassing system. In whatever paradigms or patterns are of interest to system scientists and engineers, this holds true: that an understanding of componentry is only a first step; relationships are what make the system more than the sum of its parts.

Both software projects Kahan instigated are pertinent to understanding these relationships in biological systems: Biocellion enables an empirical exploration of their implications through modeling and simulation of cellular systems; Grappa provides an efficient computational infrastructure for applications exploring big graphs. Kahan’s affiliation with ISB is made in this context: a shared belief that the exploration of multiscale phenomena is foundational to systems biology; and that in enabling this exploration, high-performance computational graph analytics, modeling and simulation tools are essential.

Parallel Algorithms
Application Tuning
High-concurrency Multi-threaded Systems
Application Performance

MA in Organizational Systems Renewal, Seattle University, 2012

PhD in Computer Science, University of Washington, 1991

BS, MS in EE, University of California, Berkeley, 1983, 1985