Rendering > Cluster
File - Contains the following file options:
New - Creates a new set of presets for clustering.
Load - Loads the presets for clustering from an .xml-file.
Import Legacy - Loads the presets for clustering from a legacy .xml file.
Save - Saves the current definitions into a cluster configuration file (.vhpc).
Save As - Saves the current definitions into a cluster configuration file (.vhpc).
Edit - Contains the following editing options:
Cluster Config - Opens the Cluster Config dialog box, where you can add multiple render nodes at once.
Configuration - Add the configuration. A wide range of formats is recognized, including the connection string from the Cluster module in previous VRED releases.
PC1 PC2 PC3 PC4
PC1, PC2, PC3, PC4
PC1 - PC4
Master1 {PC1 PC2 PC3}
Master1\[forwards=PC1|PC2|PC3\]
Master1 {C1PC01 - C1PC50} Master2 {C2PC00 - C2PC50}
Add Compute Node - Adds a rendering node to the cluster.
Remove Compute Node - Removes a rendering node from the cluster.
Add Network - Adds a network to the cluster.
Remove Network - Removes a network from the cluster.
Data Compression - Compression method used to transfer the scene graph from VRED to the cluster nodes.
Off - No compression.
Medium - Fast but low compression ratio.
High - Slow but higher compression ratio.
Image Compression - Compression method used to transfer the rendered images from the cluster nodes to VRED.
Off - No compression.
Low (lossless) - Low compression without loss of quality.
Medium (lossy) - Fixed compression ratio of 1/3 and fast.
High (lossy) - Slow but high compression ratio.
Start/Stop Action - This option can be used to prevent excessive rendering times on cluster stop or cluster crash.
Do nothing - No action is performed on cluster start or cluster stop.
Downscale Off/On - Switch downscale Off on cluster start and back to On when the cluster stops.
Raytracing On/Off - Enable raytracing during cluster rendering and use OpenSG rendering when the cluster stops.
Rendering On/Off - Rendering is switched off, when the cluster is not running.
Still Supersampling - If the CPU use during still antialiasing is poor, then by increasing the image size the use could go up. With a supersampling of 2x2, 4 pixels are rendered by the cluster and later combined to a single pixel. To get the same rendering quality, the number of image samples should be reduced by a factor of 4. (9 with 3x3, 16 with 4x4)
Homogeneous cluster - If not selected, the load balancer measures the performance of the render nodes when the cluster connection starts. If selected, the load balancer assumes that the render nodes perform the same as each other and does not measure their performance. Select this option if all render nodes use the same hardware to avoid unnecessary and possibly inaccurate measurements.
Parallel Frames - On clusters with many nodes (>100) and low-resolution rendering, use this option to increase the frame rate. If there are 500 nodes and Parallel Frames is set to 2, then 250 nodes are rendering one image while 250 are rendering the next image. This increases the latency and should only be used on high frame rates.
Frame Buffers - VRED is able to render a new image while the last images are still transferred over the network. Slow internet connections increasing the number of frame buffers could increase the frame rate.
Description - For documentation only.
Enabled - If not selected, all the nodes connected to this network will not be used during cluster rendering.
Encrypted - The whole communication in the network is encrypted. For each session a new key is used. The communication is safe against listening and recording network sessions. If you expect attacks to your network infrastructure and elaborate "man in the middle" attacks, you should consider using other mechanisms like encryption hardware or encryptions that are secured by signed certificates.
The following algorithms are used for encryption:
Key negotiation - Unified Diffie-Hellman, Ephemeral Unified Model Scheme (X9.63) or Static Unified Model (NIST). 1024 Bit
Transport layer security - AES128 Bit.
Network Type - Contains the following network type options:
Ethernet - Standard network connection
Infiniband - On linux the ibverbs interface is used to communicate with infiniband network cards. In some situations, this can be faster than using IPoIB. To be able to use Infiniband, the nodes must also be reachable over Ethernet.
Network Speed - VRED uses this option to optimize the communication. In most cases, select Auto. With 10GBit Ethernet you can get a higher throughput by selecting 10 Gigabit. If you have a 10GBit link to a 1GBit switch, then select 10G Uplink to 1G.
Service Port - This is the TCP socket port where the VREDClusterService is listening for incoming connections. Communication over this port must not be blocked by a firewall. 8889 is the default port. To start VREDClusterService on another port, use the -p option.
Name - IP address or host name of the compute node.
Enabled - If not selected, this node is not used for cluster rendering.
Use for Raytracing - If not selected, this node is not used for raytracing and no raytracing license is required.
Opens the Display Configuration dialog box, where you can set up a cluster to project (for example, on Powerwall or Cave).
Start Cluster - Starts the Cluster.
Stop Cluster - Stops the Cluster.