Totally agree with this. IMO to be useful to “less technical” artists, this is how Brenda would need to work to be useful:
-Works on Windows & Mac too
-Is an add-on or plug-in for Blender (or a stand-alone working alongside stand-alone Cycles)
-(ideally) Works as a compute device inside the rendering options (like GPU vs. CPU)
-Setup: only requires you to make your AWS account and enter the account info into a GUI
-Tells you the expected cost of a render job before starting
-Does everything “behind the scenes”… packs the scene, sends to AWS, runs the job, retrieves the files, displays the progress…
Right now Brenda is too involved for most users. But if some of the more challenging steps were taken care of by an installer/GUI then I agree that it would be a very popular render farm solution.
I’ll see if I can up a 2.7 AMI sometime this week, I’ve got a private one but it’s just a couple more steps for a public one (just deleting your credentials and traces). Oh, and it works on Mac already. I had some discrepancies between the repository provided Blender and official 2.7 so the AMI gets Blender not from the repo but straight from the foundation website.
I just finished setting up the AMI only to find out you can’t make an AMI public that has a marketplace product code associated with it. I’ll set it up again but for now I am fed up.
Had a pretty intense couple of weeks working on some projects recently. I couldn’t have done it without Brenda.
I was thinking about starting a Brenda “fan” website where people could consolidate information tips faq etc. Maybe have a wiki, a forum and something that would keep people up to date with spin-off projects. Obviously it wouldn’t work unless people contributed their experiences and knowledge. Does this sound like a good idea? I know I would find it useful.
Hovercraft: Thanks so much for the Virtualboxsuggestion. I gave it a try and got it up and running. Far more convenient. Apparently you can use that s3cmdtool with Cronand create a sync between one of your buckets and a local folder. I’m going to try it on Virtualbox.
I’m only an amateur, my pc works fine enough to model etc, but is not fine to render. Brenda is a great solution so i started testing it few days ago.
Trying the Free usage Tier and t1.micro EC instances, i was able to render 1 pic per instance, but the subframe rendering doesn’t work, or i’m not able it to work.
Anybody has got the same issue? Anybody managed it to work?
I haven’t had a chance to use Brenda, but I really appreciate the original creator and all of you sharing these tidbits. I was wanting something like this and low and behold it’s already created! I don’t have any projects that require it yet, but I hope it’s still alive and kicking when I do
@rider_rebooted, that sounds like a great idea. Please post the link and I’ll start contributing.
I’m just getting started with Brenda. Downloaded a sample render project and used it following the documentation in the Readme.md on James Yonan’s GitHub site. I am trying to figure out how to select the free storage area for testing instead of the Reduced Redundancy Storage that Brenda is using. I’ve also am looking into getting Brenda to retain performance numbers for render jobs. James tells me that he’s thinking about this feature, and I’m looking to see if I can add it.
UPDATE: Okay so the previous AMI was based on Ubuntu 14.04 and it was causing issues as Brenda could not login as root which is required. I have created a new AMI based on Ubuntu 12.04 and it seems to be working better.
Please use this AMI ID now: ami-f4c20f9c
Hey guys, well I’ve been battling with AWS all week trying to create my own publically available Brenda AMI with Blender 2.71, and I can finally say that I think I’ve done it!
The ami id is: ami-424f8c2a and in order to switch to this ami in your brenda, you have to edit the ami.py file in your brenda folder. You may have to run “python setup.py install” in your brenda folder to get it to update but I didn’t have to.
Please try it out and see if it works for you. One caveat: I’m having issues using brenda-tool ssh tail log and brenda-tool ssh update on my system with this ami. Please report if the same happens for you as I’m not sure if it’s an issue with the ami or with my system.
Hi Todd,
Thanks for taking the time to update the Brenda AMI. Does it take a lot of work to do for each update or has it become easy now you know how to do it?
Clicked on the link to check out your site but doesn’t appear to be up yet.
@rider_rebooted - No problem, it took a few days for me to figure it out, but now I could probably create a new AMI in about an hour. If you needed an AMI with a specific version of Blender that’s older than the latest weekly trunk build, it might take a little longer.
Thanks also for pointing out the broken URL, I’ve fixed that so you can go ahead and check out the site. Please let me know if there are any forum categories you think might be helpful that aren’t there currently.
I’d really like to make this site a central hub for new and experienced users of Brenda alike, so please spread the word.
That’s fantastic - just what Brenda needs. I was going to set up something similar but have been a bit busy recently moving house.
Forum category suggestions (from looking at the screenshot)…
A category where people can showcase and discuss their Brenda spin-off projects using the Brenda code.
-FAQ
It also might be cool to start a Brenda Wiki in the future.
I was thinking about doing a full beginners tutorial for using Brenda in Windows using Virtual Box, Cron etc to render and have frames automatically sync back to a Windows folder. I’ll put that up there when I get some free time.
Just looked through the github, is there a reason why they arent all rendering to the same folder & placeholder no overwrite turned on? This is the setup we use at work which allows the computers to work together, and only need to load the assets (open file / load images into ram etc.etc.etc.) once… this can save sometimes 30-50 seconds of the render, and when dealing with fast images / long sequences it can mount up.
I guess that won’t make a difference if you spawn off the same number of instances as there are frames.
Just wondering why it was designed this way instead of the no overwrite / placeholder option?
The queue is managed by brenda and the amazon service, no need for overwrite or placeholders. You have x instances rendering blocks of y amount of frames each, if one of them crashes another instance is spawned taking over the remaining block. The same system is used in all network render managers I’ve encountered.
True, but it doesn’t mean its the most efficient… We use to use that sort of setup before… where we would batch it out through a render manager… but transferring a ~300-400mb of data for each render for ~50 computers… its a lot of data for a 1GBE connection for each frame… not to mention load times and what not… most of our projects are about 3000-4000 frames long… so we woudl be talking about 400mb per frame… so about 1.5tb worth of network transfer, as opposed to 2gb.
What i was suggesting was to have a folder mounted on all of the connected amazon services & all the computers render to that location with the placeholder / no overwrite option… brenda could monitor which computer is rendering which frame, and if they crash delete that placeholder out and spawn off a new instance. it just wouldn’t handle sending out of each individual frame.
Time saved for each frame = money saved, and if i recall correctly AWS charges for network transfer aswell… so bandwidth saved = money saved.
The data is only transferred once per instance, it then spawns the blender process with something like -s block_start -e block_end. And I really wouldn’t worry about data transfer efficiency, this is AWS we are talking about, probably one of the most high-tech network infrastrucutures in existence. Network transfer is only charged between zones if I recall correctly, so make sure your bucket and instances are in the same zone.
The problem with your idea is that a bucket can only be mounted to one instance at a time so no concurrent writes to the same destination. For data input there is this
