Those of you who have used our service for a while know that we’ve had limits on how many jobs you are allowed to have rendering simultaneously. In the past, this has always been tied to Power level, which seemed like the most logical user-controlled mechanism to utilize. There’s a bit of history on why we worked this way, however.
Prior to the implementation of Harpoon, our own home-grown farm management software, our infrastructure had to operate within the limitations of the previous farm management software used to build Pixel Plow. That software had a relatively low “slice” limit in its queue. A “slice” was the smallest portion of a job, like a single frame of a frame range job or a single tile/strip of a single frame job. In order to prevent exceeding that slice limit, our software would intentionally wait to add jobs to its queue until it knew there was room for the job under the slice limit. We decided to limit the number of jobs any one user could run simultaneously to greatly aid that prevention mechanism, because when the limit was exceeded…ain’t nobody happy.
Additionally, our software had to frequently (and sometimes repeatedly) instruct that farm management software on how many CPU’s to expose to each job. While this process should have been easy, it ended up becoming more and more difficult as we grew, with sometimes hundreds of jobs in the queue simultaneously. This lead to timeouts while our software awaited an affirmative response from the farm management software indicating the CPU count had been changed. These timeouts were notably and measurably worse with more jobs in the queue, so the decision to limit per-user simultaneous job count was validated.
Now that Harpoon has supplanted the original farm management software we used, those total queue slice limits are gone. Phew! Since all of our scheduling magic is native inside of Harpoon, there is no longer a dependency on an outside application telling Harpoon how many CPU’s to expose to a job. With both of those limits eradicated from our internal workings, we have decided to relax a few things.
Today, we announce that there are no longer fundamental limits on the number of jobs you may have actively rendering on Pixel Plow. Yea! This means you may, for example, submit 30 jobs and have them all running simultaneously. We know this is more like users intuitively expect us to work, so we’re happy our technology and infrastructure can now accommodate that. Further details on simultaneous job execution, most importantly including how Power level is handled, can be found on this page.
We’re happy to help you by allowing you to do more on Pixel Plow. We continue to work on ways and methods to make our service the best render farm service available. So go ahead and render like crazy!