Remote OpenGL Performance

OpenGL Development JOGL Development
#1

Hey all,

Can anyone help put in terms what sort of network connection and anything else I’m not seeing that it would take to render models w/ JOGL over a LAN from one building to another so that there is jerkiness at all in the rotation control and that it looks just as smooth as if it were running locally or what would have to be done to get to this point.

I’m working on the pro’s and con’s of doing this with my app and I’d have to get a package that does OpenGL extensions for my X Server (Exceed). Currently, the network is a 10/100 Ether w/ a few firewalls in between.

Any advice/comments much appreciated.

#2

In theory an OpenGL accelerated X server over 100 Mbps Ethernet and relatively few hops (so low latency) should give you pretty good interactive performance. Some eight years ago my former group did some experiments and interactive performance over 100 Mbps to Exceed (as I recall) was pretty good.

Today though my recommendation would be to utilize Java to run the client locally. Is there a reason this is not an option?

#3

It’s just something I don’t want to have to deal with. Having to budget money to buy new Exceed & 3D packages, then rearchitect my app so that it works correctly over the network.

Some desktops that I run my app on don’t even render it correctly directly, not to mention indirectly. And then there are the network issues of latency and bandwidth with a dynamically generated model.

And there are even GLX calls that don’t work over the network or w/o hardware acceleration.

My app just wasn’t built for this so it’s kind of a pain when someone says “Let’s centralize our servers”. Even though I make no use of this cluster in a server like fashion. My app is sort of a front end to a cluster, like a workstation console.

I was just wondering what other issues I’m going to come up against that maybe someone else has that will completely stop me in my tracks.

#4

My point is that if you are dealing with Windows clients you would probably be better off doing your own networking and running a local Java client talking to your cluster rather than running an X server like Exceed and hoping the GLX acceleration works. You basically seem to be in agreement with this which is why I’m a little surprised you’re thinking of going this route.