Skip to main content
P
ptc-47985091-Visitor
1-Visitor
June 18, 2014
Question

XenDesktop and Creo Parametric 2.0

  • June 18, 2014
  • 4 replies
  • 1312 views

We are evaluating the use of Citrix XenDesktop for remote users of Creo Parametric 2.0. Although it installs just fine, there are issues with the video settings.


Any assistance will be greatly appreciated as we have only a short time left on our proof of concept installation.

4 replies

P
ptc-19327451-Visitor
1-Visitor
June 18, 2014
We have just demo'ed this with no problems whatsoever using NVIDIA GRID cards on the server side. No video problems that I can see. We only looked at single display installations so far though. Production will most likely be at least 2 monitors per user.

Runs CREO2 M100(?), ANSYS V15, Vericut 7.3, etc...very well. Server was offsite and still ran well.

I know we had two different horsepower NVIDIA GRID cards to choose from, we chose the higher-end to test.

If you find this at all helpful, I can put you in contact with our IT person who can describe the setup better than I can.


Christopher F. Gosnell

FPD Company
124 Hidden Valley Road
McMurray, PA 15317
E
egifford12-Amethyst
12-Amethyst
December 17, 2014

Chris,


Has your company moved into production use of running Creo 2 etc remotelyon hardwarethe datacenter? What was the client configuration you tested? Did you just use a regular PC and connect to Citrix or were you using hardware like Dell'sWyse lightweight clients?I want to start exploring options of using Citrix and hardware like the Nvidia Grid cards for a possible future use case, so anything I can learn from actual implementations is helpful.



Thanks


Regards,


Erik Gifford

P
ptc-19327451-Visitor
1-Visitor
December 18, 2014
We have the grid in our capital budget for 2015. We are pretty much sold on the technology and our testing was very successful for Design and Manufacturing. We do not use PDM.

We are looking at 3 Xen servers, each supporting 8 clients. We will use two for active work with one for overflow for approx 12-16 users. We planned on running the Citrix software on our existing 4 year old Dell-5500 dual core workstations, with the option of buying new hardware in 2016 or 2017. The focus then (if the Grid works as we expect) will be replace current workstations with laptops or even portable tablet style computers that can connect to high resolution monitors, mouse, etc... when docked. We are really looking forward to having our Engineers be able to walk around out plant, or even go offsite with their full complement of apps available to them.

Using this scenario, the cost is a wash as compared with just buying new workstations during this upgrade, but we expect the future hardware cost to decrease dramatically in the near future.

My personal interest is to ask (push) my software vendors to improve their software to make use of multiple cores.
I don't know how long PTC can hold out without multi-core parallel capability. As a point of interest, I just viewed a demo for Autodesk HSM machining that makes use of up to 36 cores!

I understand that using the NVIDIA Tesla API's are proprietary, and some vendors are not ready to do this as compared with the open-source MPI type libraries. I would hope that in the near future someone creates a 'wrapper' for MPI calls to Tesla drivers to make existing software easier to port and less risky.

Christopher F. Gosnell

FPD Company
124 Hidden Valley Road
McMurray, PA 15317
K
kimndave91-Visitor
1-Visitor
December 19, 2014

To be fair(er) to PTC, the multicore support in HSM doesn't seem to apply to the CAD application but only to the CNC path generation segment that was purchased as part of Autodesk buying HSM from under Dassult's reach.They also support splitting the NC path generation to entirely different machines. It's relatively easy to do for NC path generation. Calculating the roughing pass and the final pass for each feature can be done independently. Creating paths for drilling and tapping doesn't depend on milling calculations. Paths for separate sides of the parts are also independent.


It's an interesting problem as described here http://www.3dcadworld.com/solving-cad-concurrency-problem/


In the comments on concurrency a statement is made that D-Cubed is multicore capable (it says version 58 (2D), but I found a ref in 47(3D)), which the D-Cubed site http://www.plm.automation.siemens.com/en_us/products/open/d-cubed/product_news/3ddcm_version47.shtml indicates that it is only thread safe, not managed distribution."From this version 3D DCM creates a series of independent discrete tasks, each representing a portion of the requested calculation. These tasks are passed to the application which can then run them using any desired parallelization technique."


In other words, if the running application can figure out some way to manage the parallel tasks, D-Cubed routines can be dispatched to separate cores. It still leaves scheduling and setup of those tasks to the host software. Scheduling and setup are the hard part of parallel execution.



In looking through release notes, it's interesting to see the sort of controls that D-Cubed offers software developers. I liked the one on chirality controls. These prevent the obnoxious flipping between solutions by adding more refined controls to detect a change between shapes that otherwise meet explicit geometric constraints. http://www.plm.automation.siemens.com/en_us/products/open/d-cubed/product_news/2ddcm_version62.shtml

Terms & ConditionsCookie settingsAccessibility statement