[Maxer]

That gives me a little hope since all the machines rendering are on Windows XP64 and the manager is on Windows XP.

[Mihai]

Maxer and b-kandor, what systems and Maxwell versions are you using? On your manager and servers respectively.

[Maxer]

On the manager I'm using Windows XP with Maxwell 32, on the servers I'm using Windows XP64 and Maxwell 64.

[b-kandor]

The examples I posted (where the final merged mxi ends up being in the temp folder of the manager system) where rendered with the following:

1: winxp64 with MR1.5 running manager, server and mxcl
2: winxp64 with MR1.5 running server
3: winxp32 with MR1.5 running server
4: winxp32 with MR1.5 running server

Kandor

[Maxer]

I guess I should say that they were all V1.5. I've had to go back to V1.1 so I won't be able to test anything now.

[tom]

My config for this coop test was:

alpha - x64sp2 + mxcl64 (manager+server+viewer)
filmboy - x64sp2 + mxcl64 (server)
titan - xp32sp2 + mxcl32 (server)
hurricane - xp32sp2 + mxcl32 (server)

See video here:
http://www.divshare.com/download/1000450-cbb

[Q2]

THAT's BEAUTIFULL.

I just wish it would work like that here on my machines. Wired thing is that I do exactly the same but the servers keep crashing or the distribution to the servers hangs forever at 20%.

BUMMER!

Q! Berlin

[Maxer]

Tom I watched your test and I think it's important to tell you that yes I believe cooperative rendering is working but it's not working properly. In your test you ran it for about two minutes and reached a combined SL level of 6.61 and the highest level any of your single machines reached was 6 so it's obviously working. In my case when I ran it none of my machines reached a SL over 12.xx but the combined SL was 14.xx so it's obviously combining the images to get a higher SL. The problem for me comes in when I compared my fastest single machine against the 11 machines and found that the single machine was reaching a higher SL in the same period of time as my 11 cooperative rendering machines. I can't be to clear here, yes cooperative rendering is combining the separate MXI's in cooperative mode in order to get a higher SL but the resulting SL isn't as high as it should be.

I ran a test with 1.1 last night for about 10 minutes using the same machines and the same setup as I did with the 1.5 test. After 10 minutes the combined SL from 10 machines was 12.17, my single machine which is a dual processor quad core system reached a SL of 10.02. I suggest you run the same test that you did with 1.5 with 1.1 and see if you get the same combined SL for each job. I can almost guarantee you that the combined SL from 1.1 is going to be higher than it is in 1.5.

[tom]

Maxer, I agree. Today, I could reproduce the problem which also b-kandor has mentioned. It seems like somehow sometimes (I'm not yet sure if always..will keep testing, it's not easy.) the merging process cannot merge all the temporary MXI. If found one of my nodes was merging while other was merging and writing the final MXI without waiting for the other merge. As a result SL 6.88 was written as a final MXI and others had merged another as SL 7.70 which had to end up as SL 8.97 which explains all the performance problems. As a conclusion, I am pretty sure every node renders fine by looking at the temp files but it seems like there may be problems with merging process. I will repeat this test with all 32 bits and see if there's also problem. Meanwhile, for the best result without problems accept my honest suggestion. Do coop and don't stop from the job tree. Stop each node separately and merge their cooperative.mxi (not _aux) files manually. I know it's frustrating but thus, you can be sure about the performance until this is fixed. Thanks for your efforts.

[simmsimaging]

how do you stop each node? Is there command or do you just shutdown the server node?

b

[tom]

Select one of them and hit stop.

[simmsimaging]

Oh - I get it - from the mxcl window, just not at the top of the job tree. Sorry, I didn't understand that at first.

b

[Maxer]

Thanks Tom I'm glad that we are figuring out what is wrong and I hope it gets fixed soon. Your suggestion is good but it's not practical when dealing with more than a handful of machines so I'm sticking with V1.1 until this is fixed.

[tom]

Thanks for all your help. As long as we can identify the problem -which means translating complaints into technical trace reports- we can find and kill the bugs. We'll keep you updated about this.

[b-kandor]

Hi Tom,

I just wanted to add this information:

Another symptom I've experienced is when displaying the in-progress image in viewer (when it takes a minute to compute the image), then you click on the new tab created to view it.

Sometimes the sl shown for the 'temporary merged image' is lower than the sl of my fastest machine and sometimes it isn't. But, if it displays lower on one job it will always display lower on that job and vice-versa.

Kandor