FAQ — Pixar's Tractor™

1. What is Tractor™ v1.0?
   Tractor is Pixar's new industrial-strength distributed computing solution built on an architecture designed to handle the largest multi-core render farms of tens of thousands of cores or more.  Tractor is designed to greatly simplify deployment, configuration, and control of rendering assets while emphasizing scalability and performance. It will accept interactive and batch jobs from all users on the network, analyze priorities and hierarchical task dependencies, and dispatch work across the render farm.  Any site-specified command can be dispatched from rendering, simulation, and compositing, to other computational or administrative activities.
2. What will Tractor™ v1.0 not do?
   Tractor is not an asset management system and it does not address the logistics of deploying and configuring actual rendering assets. (Scene, geometry, texture files, etc) Tractor’s primary focus is scalability and ease of maintenance.
3. What are the main functional components of Tractor™?
   Tractor is a distributed computing solution designed to accommodate the very large multi-core networks that are planned or already starting to appear at production facilities around the world. Tractor provides significant scalability over current generation solutions, such as Alfred, through a combination of high-performance multi-threaded assignment logic and a streamlined overall architecture.  There are three main Tractor components:
   — 1) Tractor Engine, the central job queue, task assigner and status monitor.
   — 2) Tractor Blade, the remote execution server on each farm host.
   — 3) Tractor Dashboard, a browser-based GUI designed around an extensible framework.
   Tractor uses open and extensible web standards as a fundamental feature of its implementation.
4. What is the development focus for Tractor™ v1.0 and what can be expected at Tractor v2.0?
   We have big plans for Tractor and the development schedule for v2.0 and beyond is already busy. The v1.0 release will be focused on the fundamentals of a scalable job queue and assignment engine. The user-interface for v1.0 is simple, functional, and provides an accessible and extensible platform, however, there are some configuration and management functions that are explicitly not addressed by the v1.0 GUI that will be addressed at v2.0.  At v1.0 these functions can be addressed using simple text editors, or even via site-defined scripts that take advantage of Tractor's HTTP and JSON interfaces. Tractor does not have built-in support for SQL database operations; however, data transfers are possible through a variety of other approaches. 
5. When will Tractor™ v1.0 be released?
   Tractor is now in beta and release is currently scheduled for later in 2009. (Subject to change)
6. Can I join the Tractor™ v1.0 beta program?
   We are always looking for beta candidates who can stress test Tractor. Given the development focus of Tractor v1.0 outlined above, if you believe that you can offer us quality feedback please contact us at rendermansales@pixar.com.
7. How does Tractor™ differ from Alfred™?
   Though Tractor functions in many ways as a “next-generation” Alfred, it is also intended to be considerably more than Alfred in terms of scalability for very large farms. Tractor also strives for a greater industrial simplicity in its architecture and baseline assumptions, resulting from our analysis about the way large, modern render farms are now developing.
8. Will Tractor™ be licensed in the same way as Alfred™?
   No, we have adopted a total capacity licensing approach for Tractor.  Unlike the distinct Alfred dispatcher and per-node Alfserver licenses, Tractor will simply be licensed by the total number of concurrently active tasks on the farm, independent of the number of remote hosts. Tractor-Blade servers can also be started at host boot time and remain running without active connections to the license. The valuable flexibility offered by this "total concurrent capacity" capability ensures that you can select licensing in proportion to your usage, from a few tractor nodes on a farm to many tractor tasks running on each node.
9. How much will Tractor™ v1.0 cost?
   As Tractor is implemented and licensed differently from Alfred it will be released with a new and competitive pricing model. Please regularly check this website or contact rendermansales@pixar.com when prices will be announced later in 2009.
10. Will existing Alfred™ users receive Tractor™ v1.0 as a free upgrade?
    Yes, RenderMan Studio and Alfred customers on current maintenance will receive a direct conversion to Tractor over a two-year transition period that will commence when Tractor v1.0 is released. An upgrade price will be available for those RenderMan Studio and Alfred customers not on current maintenance.
11. Is Tractor™ designed only for RenderMan or will it work with any product?
    Though render management is the focus of Tractor v1.0, the architecture of Tractor has been specifically designed as a generic solution for a multitude of distributed computing tasks. Whether you wish to manage compositing, simulation, or any other task, Tractor is designed to address these requirements.
12. Will Tractor™ be available as a standalone product?
    Yes, Tractor will be competitively priced as a standalone product. Please check back at this website or contact rendermansales@pixar.com where prices will be announced later in 2009.
13. Q. What platforms will Tractor™ v1.0 support?
    Job spooling from users, and task execution on nodes will be supported on Linux (32 and 64-bit), Mac OS X, and Windows (XP, Vista, 7) when Tractor 1.0 is released.  The central job queue engine will initially be available for Linux and Mac OS X only, with Windows releases also planned.  Please check back at this website in the fall of 2009 or contact rendermansales@pixar.com for further information.

If this FAQ does not completely answer your questions, please contact us directly at rendermansales@pixar.com.