Each job task can be tracked in real time
Alfred is the ideal foreman for your renderfarm. Alfred is a task-based processing system specifically designed to manage distributed rendering. It manages the render tasks of an entire studio efficiently, automatically, and robustly. Alfred is tightly integrated with RenderMan Studio, and supports a powerful scripting language that allows integration with distributed computing requirements at your site. An optional remote execution server, Alfserver, collaborates with Alfred for added scalability. Alfserver adds significant capabilities for launching and tracking remote commands – a big win for larger renderfarms. Trust Alfred to automate your Studio’s render pipeline.
Alfred manages resources by working from a “white board” that you control. It identifies system resources including rendering machines, groups, and individual users, and specifies the times they’re available. It controls rendering servers from a shared pool as needed. If they’re all in use, it waits until an appropriate machine becomes available. Alfred simplifies system administration by using a pattern-matching scheme to make service requests, so specific host names are unnecessary. System administrators can “hot-swap” machines from the pool even as jobs are executing. Coupled with user-defined priorities and access times, this approach ensures the highest possible productivity. Alfred’s command language adds still more power. Use it to distribute the separable rendering components of a final frame (such as shadows or reflections) to remote network renderers in parallel. You’ll get the fastest single-frame rendering possible and take full advantage of the computing powers of your network.
Alfred ships with Studio, and is optimized to seamlessly manage the application of RenderMan assets to your Maya files. You can develop your own “Alfred savvy” applications, and even create distributed services for Alfred to manage. Alfred provides progress notification, remote message collection, and automatic restart of tasks that have stopped due to errors or even system crashes.
With Alfred, you control your own jobs directly from your individual machine, instead of via a central batch processor. So multiple jobs can be active at the same time throughout the network, making optimal use of network processors and disks, with user permissions guaranteed. This arrangement adds fault tolerance as well. Since it doesn’t do any job processing itself, Alfred imposes only a modest demand on your local CPU.
Although Alfred’s components are general purpose, together they make networked rendering of RenderMan effects extremely efficient and easy to administer. Alfred consists of these independent parts:
Control jobs in the dispatcher
Alfserver is the server-side component of this job distribution system. It's Alfservers job to monitor machine health, readiness, and any other arbitrary metric you want. Alfserver is include with the package, RenderMan Pro Server.
TM & © 1986 - Pixar. All Rights Reserved.