IDOCs with SPHs slower rendering than Full SPH!

[mpowers10000]

I tried rendering a scene with five IDOCs filled with an SPH per IDOC. I rendered using five nodes on renderfarm, but was significantly slower than rendering a Full SPH!

I tried this with SPLASH, BUBBLES, and FOAM - each one at a time from a cached Hybrido domain. The IDOCs are scale on non-uniformly to fit simulation area.

Should I post an XML?

Thnx

[Alex]

Hi,

can we have a look at the setup of that scene? (an XML file would be ok).

We don't fully understand what do you mean by SPH per IDOC vs Full SPH.

Thank you

[mpowers10000]

By SPH, I meant secondary particles: HY_Splash, HY_Foam, etc.
Full SPH, I mean, e.g.: A single HY_Splash for the whole HY_Domain

How do I add attachments to this forum? I will send XML, as it will be clearer than my incorrect nomenclature.

Thank you.

[Alex]

Hi again,

I think you should be able to attach files from know on. Let us know otherwise.

Does the computer with the single (whole) splash has local access to the domain cached files?
Or is it reading the cache from a network drive?

The same question applies to the other elements. Are the simulations being read/written on local or remote drives?
Maybe simultaneous access is creating a bottleneck.

[mpowers10000]

1. Disk Access: I am using a local 1TB Evo 850 SSD. Our Spectra Logic Verde NAS is not functioning correctly - and our IT departments EMC VM shares only have ~230 MB/s throughput. (I am getting ~ same throughput to the render machines from my local SSD.)

However, I have tried simming the single secondary particles on a farm computer and it is still much faster. Perhaps I have connected the IDOCs incorrectly.

I am sorry to be so obtuse, but how does one attach a file on this board? (I don't see anything on the dashboard related to attachments)

Thank you

[Alex]

No worries, it's our fault. I believe finally you should be able to post attachments.
There should be a second tab below the Preview/Submit buttons called "Upload attachment".

At worst case, maybe you can post its content inside a

Code: Select all

 block, but I don't know if it would exceed the post length limit.

The tests you are doing seem to be ok. Maybe some of the secondary elements of the IDOCs is not actually set to bounded, and still keeping everything into account.

Hopefully we will be able to confirm it with the XML file.

Sorry for the inconvenience.

[Alex]

Hi again,

thanks for sending me the XML file.
Sorry for all the obstacles. You also helped us to fine tune the permissions (it should be possible to attach XML files from now on).

Back to the initial issue. After checking out your scene, this is what I have seen so far:

• Splash nodes seem to be properly configured (bounded).
• Bubbles nodes are not set to be bounded. This means that every bubbles node is taking into account the whole domain. If you created the nodes using the "Bubbles per IDOC" command, this is something that we must fix.
• Foam nodes are not set to bounded. This is not wrong, depending on what you really want to do. Foam nodes have enabled the "From domain" parameter. Again, this means that the whole domain will be taken into account.
  Not using the "Bounded" option makes sense for foam created from the splash, because splash particles may fall out of its original bounding box. However, if "From domain" is enabled, you will need to set it to bounded if you want to avoid the extra computation.

Hope it helps

[mpowers10000]

Thank you for the excellent answer!

I will adjust those values. I ran those sims last week, I'll revisit them and run tests of the farm.

I did create the bubbles with Bubbles per IDOC command


Thank you, Alex