Dr Michael Hobbs

1

• Dr Michael Hobbs
• School of Engineering IT
• Faculty of Science Technology
• Email mick_at_deakin.edu.au

2
Overview

• Introduction
• Rendering Examples
• STALGS Project
• Design Implementation
• Partnerships
• Conclusion Questions

3
Introduction

• The goal of this project is to research and
  develop a system for the rendering of digital
  media outcomes
• Rendering is the process of generating an image
  (still, animation or movie) from a3 dimensional
  (3D) model of a scene or objects, using
  specialised software

4
Introduction

• Rendering is a computationally expensive
  (consuming) process
• But, it is inherently parallelizable!
• Meaning, we can chop the job up into smaller
  pieces and execute them concurrently
• Distribute these smaller jobs over a set of
  distributed, or cluster of computers
• A cluster of computers used for rendering is
  termed a Render Farm

5
Examples 3D Models
6
Examples Architecture
7
Examples Architecture
Note This 800x600 single frame image took over
35 minutes on my desktop PC
8
Examples Motion Capture/Animation
Act3 Animation Render
9
Examples Games
10
Examples Multimedia
11
STALGS Project History

• Many areas/courses within the University require
  the creation of high quality, digital images
• Faculty of Arts are actively developing new
  courses in animation Motion Capture Lab
• Faculty of Science and Technology with new and
  existing courses in Games Design, Multimedia and
  Architecture
• The needs have grown beyond the Universities
  resource capacity within students labs

12
STALGS Project History

• Traditional rendering is
• Time consuming
• For the student and teacher
• Computationally expensive
• Ties up student or lab machines for
  hours/days/weeks!
• Limits the students development and exploration
  opportunities
• Students not experiencing current technologies
  used in the industry

13
STALGS Project History

• These issues were being seen as a constraint on
  the teaching of these various course
• It was seen that together we may be able to
  provide a solution to these issues
• A shared Render Farm resource
• Submit a STALGS Proposal!

14
STALGS Project Team Members

• Leinors Torre (Project Leader)
• School of Communication and Creative ArtsFaculty
  of Arts
• Chris Welsh
• Knowledge Media Division
• Michael Hobbs
• School of Engineering and Information
  TechnologyFaculty of Science and Technology
• Kim Vincs
• School of Communication and Creative ArtsFaculty
  of Arts
• Sambit Datta
• School of Architecture and BuildingFaculty of
  Science and Technology

15
STALGS Project Goals
Teaching
Fast, high quality visual images and
animations Faster development and experimentation
for students
Industry/Service
Research
Many interesting research problems web-based
interface, load balancing, rendering algorithms
Potential cooperation with other
institutions/organizations Possible commercial
resource
16
Design Requirements

• It is a requirement that this service be
  accessible by a variety of student groups within
  the university
• There are a variety of software packages being
  used to create the models
• Maya, 3D Studio Max, AutoCAD, Belnder3D,
• Thus any solution should be widely accessible
• both in terms of location and acceptable formats

17
Design Process

• Examine existing current technologies and
  solutions
• Test environment Proof of concept
• Our objective is to quickly establish critical
  mass of client nodes
• DrQueue with Maya
• Virtualalised client nodes- Mandriva Linux
  running on VMWare server
• Real servers - Mandriva on Dell 2850 dual quad
  core servers

18
Design - Web Interface

• This issue - Submission and management of jobs
  from anywhere
• Parity - Off-campus and on-campus students
• Ease of access. client needs to be operating
  system independent and accessible via DSO
• Network speed - submission of jobs for off campus
  students
• Management via Web Interface has advantages but
  does not solve all the issues

19
Implementation

• 2 x Dell 2950 servers
• Maya Rendering Software
• DrQueue Job Management

20
Partnerships

• There are a number of ways this project is using
  partnerships to inform teaching and learning
  developments
• Partnerships between staff from different
  faculties is expanding the possibilities for
  joint projects and studies
• Partnerships between students from different
  faculties is nurturing the real-world scenario

21
Partnerships

• Collaboration is already occurs between the
  Faculty of Arts and Faculty of Science and
  Technology
• Combining/sharing computational resources
• Combined student projects
• Future collaboration is being negotiated with
  VPAC

22
Conclusion

• This STALGS project is addressing an important
  problem
• The key design issues for such a Render Farm have
  been completed
• A pilot Render Farm system has been installed and
  is being used by staff and students
• Preparation to test this on a larger scale is
  being done

23
Thanks

• Any questions?
• Dr Michael Hobbs
• School of Engineering ITFaculty of Science
  Technology
• Email mick_at_deakin.edu.au

24
(No Transcript)
25
STALGS Project Goals

• Focus
• Marketing
• The Motion Capture Experience
• Research
• Teaching and Learning
• Commercial

26
Motivators - Marketing

• Marketing
• Raise the University profile as a leading edge
  provider of education for high end digital
  animation
• Its important for Faculties to promote
  portfolios. (i.e. Mocap, etc). Doing so will
  attract quality students

27
Motivators - Research

• Research
• The University has poor centralized research
  infrastructure - We need to improve it
• Promote the development of infrastructure for
  research

28
Motivators -TL

• Teaching and Learning
• Ultimately provide a flexible service for more
  than one Faculty.
• Rendering high quality images can be a slow and
  an iterative process. Students can often take
  more than 24 hours to render one project in the
  labs. Often they need to render them more than
  once.
• Given the current resource limitations
  compromises have to be taken. This reduces time
  but compromises the quality of the resulting
  product.

29
Motivators -TL Cont...

• Teaching and Learning Continued
• Rendering high quality images can be a slow and
  an iterative process. Students can often take
  more than 24 hours to render one project in the
  labs. Often they need to render them more than
  once.
• Given the current resource limitations
  compromises have to be taken. This reduces time
  but compromises the quality of the resulting
  product.