High-Throughput Computing (HTC) for Parametric Exploration by SOM Sebastian Claussnitzer, Neil Katz, Matthew Shaxted, Sejung Kat Park, Robert Yori

Parametric thinking and generating an appropriate design solution space with a large set of design parameters have become the norm in our generation, especially for the caliber and scale of designs we produce in advanced practice. What has also become a necessity today is the designer’s understanding of energy usage and environmental impact of their proposed designs early on, during the initial stages of parametric exploration. Inevitably, we have been defining the limits of our explorations (number of parameters, scale of parameters or the resolution of testing model) within the scope of computation power of our machines or analysis tools of our choice. This workshop will teach and enable participants to utilize distributed computing protocols as a generative design driver for creating and subsequently evaluating large catalogues of building design options. Compared to current desktop approach, optioneering at this scale and speed will allow designers to exhaustively model the solution space for design problems, as well as gain a deeper understanding of the range of implications associated with any modelled/simulated solution. 

A relative absence of advanced decision-making processes limits the number of building-design options that designers can examine for energy efficiency. Large design spaces and limited computational resources have encouraged designers to rely on optimization techniques. Optimization alone may expose optimally performing design, however, may fail to unearth a large part of a solution which can be uncovered by a wider search through this space. Exposing these solutions allows a designer to evaluate trade-offs among the various dimensions of governing factors such as energy-optimal design, adherence to conceptual idea, or visual ethos. With the rapid decrease of computing cost and easily accessible parallel computing technologies we are proposing a new paradigm for design explorations: Using parallel computing to create exhaustive option catalogues for decision making in the design process. Those options can then be searched and graphed against many types of criteria, both quantitative and qualitative. 

Participants are expected to have knowledge of the following concepts and software applications: 
Rhino & Grasshopper 
Conceptual knowledge of Energy Modeling 
Excel 
Basic programming skills 

Preferred but not necessary: 
Knowledge of energy modeling software, preferably Energy Plus 
Tableau 
R 
Radiance