Next week is roadmap week - i’m going up to WPG to walk everyone through all the products and what we are doing next.
anything that is screaming for attention: now is your chance 
cb
New products or updates to existing ones? Or both?
both. talking about deadline 6, frost 2, krak 2.x or 3 and others.
Deadline v6 Roadmap suggestions:
-
Remove necessity to re-install repository/client between major version releases. Can we detach the auto-updating system, so its always installed on a client and the repository and then you can dump a new major release version of Deadline into the repository and it gets deployed straight away. Your licensing system will still be good, as the Thinkbox licensing system is locked to the slave. Then, if you want to release 4 major version releases of Deadline in 1 year. No problem. I’ll be there with you, always up to date, reporting bugs!
Bonus point. Crack software. ‘In my opinion’ - It’s one thing to try and ‘de-value crack’ software by delivering to market a ‘lite’ version of say, Deadline, but I really don’t think this works well for your kind of software. I much prefer the idea of smaller AGILE iterations of your software. Use Team Foundation Server to pump out nightly builds, which I can dump onto my auto-updating repository, combined with regular license updates, to slow down the torrent/crack kids. -
Pulse HA / Remote studio. I have multiple studios. I want my remote artists to have the same experience with Deadline as my local artists. Each studio has its own farm, one bigger than the other. I want my remote studio to have good access to the larger farm. I can handle the big data issues, with expensive hardware/software solutions, but Deadline isn’t there yet. Mentioned this before, but the concept of multiple Pulse servers sharing cached info, would give a seamless experience, no matter where you are? Or doc DB based setup and then use clustered backend?
-
Developer IDE Intellisense Extension / Plugin for Visual Studio and/or Eclipse IDE for the Deadline API. Me wants it bad. Ryan wants it and I reckon Lukas and Gavin would be up for it.
-
Fix all UI issues by going for a platform agnostic UI toolkit. The less 3rd party dependencies the better. Aim high for a super modern, sexy/slick UI that kicks butt. As non-Windows looking as possible!
-
TreeView, Job Dependency graph action. I want to see my job dependencies in a clear, fast manner and manipulate them in smart ways.
-
Deadline Cloud Rendering. I would much rather spend my money on bandwidth (to the cloud) & licensing AND not hardware (expensive farms). Leaving the remainder of my cash for fast local (network) storage and fancy things, like gateway accelerators which are presently rocking my world! - averesystems.com/Products_FXT.aspx
-
Renderer DBR Plugin system. Develop a system whereby Deadline can support Mental Ray / VRay DBR across the slaves. There you go Ryan. I bet you were wondering how long would it be before my mentioned this old beast?!
My projects have a life-cycle. We always have the ‘lighting’ phase of a project, before final rendering. Yet, I don’t have a means to use my expensive farm for the test lighting phase. Just local rendering, whilst my CPU and GPU cores sit idle, until we slam them for the final/full-frame test renders. Using VRay/Mental DBR to get the look/dev work quicker would allow us to stagger the final rendering slam on the farm better. -
“Maintenance Plugin” Job Type. A new type of Deadline job which executes only on a whitelist of slaves BUT for each slave that completes successfully, it blacklists itself from ever running on that slave again (unless re-queued). I would like to submit a job to X machines and execute a script, python, vbs, fire a executable, anything to each of these machines.
-
Enhanced Pop-up Handling Framework. Ability to interrogate ALL the controls of actual popup dialog as they appear. This would allow support to monitor “sub” jobs such as FumeFX simulations and update the Task Status Progress correctly. A job sitting at 0% progress after 10 hrs isn’t very helpful for feedback! Better “Sims” based job handling. Realflow, FumeFX, Naiad + add Houndini sim job support, including distributed bucket support?
-
Colourized Messages in the Log Reports. Now we have themes, ‘colourising’ the text in error/log/re-queue messages? ie: “Error” (red), “Warning” (yellow), “Information” (green), and “Debug” (grey). Coloured in the slave window and Pulse window as well? Tie it into colour themes, so user can control colour scheme of log reports windows! As a default, I would suggest leaving all INFO messages as they are, but colour up WARNING as yellow/amber and ERROR as red. This would make it really easy to see these important messages in the log/error/re-queue reports, as I think they tend to get missed sometimes…
-
Search History in the search windows of the deadline monitor & repository. Global repository setting controlling number of search results to display. I see this as a drop-down list in each search history field in Deadline which simply remembers my previous X number of searches. I find myself carrying out the same searches all the time. This would make it quicker and also act as a reminder.
-
Post image file validation - auto check that I don’t have any < x Kb files created. Therefore, corrupt, re-queue for me automatically. Ability to check individual channels/render passes and re-queue automatically would be mega-awesome. Also, ability to auto-calculate what the average frame size should be and therefore know that a certain frame doesn’t meet that standard and hence re-queue automatically.
-
Deadline Mobile - Control actual jobs in the queue? iPad edition, allowing control and graphing/charting?
-
Global farm reporting system. Ability to generate reports, almost like a SQL query. Fully customisable. ie: If I get asked to generate a report based on certain jobs all for a certain project. “How many rendering hours were used/wasted on project X” type of question. Ability to pull out images based on the graphing would be most useful for management. Helps me justify more render nodes, which equals more Deadline licenses!
-
I think Bobo has threatened a custom plugin for Deadline (bottom of one of his recent Deadline tutorials) to batch process X 3dsMax files against X MaxScripts. Yes, please! Awesome pipeline tool.
-
Custom tab in SMTD. Bobo knows all about this. Bespoke studio functions can then go in this “studio_functions_private.ms” type MXS file. Tricky bit is for Bobo to solve the best implementation of this so it can play/talk nice with the rest of SMTD. ie: I reckon we need to have ‘different’ #classes of functions in this system. #preSMTDSubmit, #postSMTDSubmit callbacks? to allow my custom functions to execute at the right time. I might want a function which affects the submitted Max file BUT NOT the resident Max file open on my machine when I hit the submit button. I think this is a good challenge and I’m excited to see what Bobo can build for us here
-
Ability to ‘near’ real-time stream the STDOUT/STDERR slave output for a particular frame to a workstation for review/rendering DEBUG would be useful. Limit this function to just 1 slave to 1 workstation to limit bandwidth.
-
Ability to have highly detailed email reporting system from Deadline. ie: per user setting, which controls the level of verbose emails being sent to the user account’s email address. ie: some of my users would like to receive the STDOUT / STDERR report logs. Auto attaching to the email as “attached_task_logs”, “attached_error_logs”, “attached_requeue_logs” zip files would be excellent. Ability to control which if any of these attached zip files get attached would be good as well. For bonus points, inserting into the actual email, the top 5 most popular error reports for that particular job, last 50 lines of STDOUT/ERR for “failed” email notifications would be very impressive!
-
Make SMTD Sanity Checker MXS code multithreaded using .Net background workers.
-
Invite advisory board members to your roadmap meetings including continuous free cocktails.
Mike
#1 Request:
Computer Configuration:
- Installation Macros
- Update Macros
- Synchronize Fonts, plugins, license inis etc.
Push out Nuke to all of our render nodes. Attach a new render node have it automatically install the latest Nuke.
Install windows. Plugin to network. Install deadline. Have it autodetect /configure repository. And then have it check what’s installed and update accordingly.
#2 Request:
Smart Deadline Job Completion Estimation
#3 Request:
Ability to “sneak a peak” at VFBs rendering as a service.
#4 Request:
MySQL/Postegre based repository.
#5
SLAM in Fathom
#6
Better UI framework for deadline
#100 Far-out-there Request:
Deadline Service.
Ability to ‘lease’ our farm out and rent unused capacity through Thinkbox as a reseller of CPU cycles. Have an encrypted backend that securely uses our farm to earn extra income.
+1 For “Run Once per slave” job.
I agree on the rendering market. I think it would be so much more efficient, and it would be easier to scale up and easier to pay for the equipment we already have. But I also like all of the stuff that would have to be developed to make it work, like VM’s, application deployment, maintenance jobs, new caching techniques, etc.
For Frost…
How about some sampling/meshing options for surface points, like what we’d get from LIDAR and other range scanning systems. What we have now works great when the points represent a blob in a volume, like SPH fluids, but for surface scans, it’s too blobby and doesn’t handle sparse sampling. Maybe check out cs.jhu.edu/~bolitho/Research … nRecon.pdf. There’s a lot of work done on this subject for a long time, so while these stand out to me, you may find better.
Maybe other inputs, like depth maps or voxels?
How about KCM’s, but for meshes?
For Krakatoa…
PRT changes. Include multiple streams to speed decompression, but also split for proxy or spatially so we only have to decompress a portion of the file if we need to get the points from one area. Maybe include neighbor lists or other things that can be pre-computed to speed things along for later? More things that aren’t per-particle, too; sorting vector and other metadata have been talked about before.
Raymarching as a normal Krakatoa rendering mode (not just the atmospheric) plus more access to voxel.
Shading KCM? Splatting KCM?
Anything that helps with iterating in a KCM.
Intermediate caching stages. Even stupid ones like caching the PRT’s to the local drives.
Sorry for being so late on this. Darn forum blew away my previous post and I was too tired to retype everything.
- Chad