Most grid uses these days are "compute" grids. As we start to look at some other interesting concepts of "grid" usage one that has been mentioned has been the "display grid". However, as I think about this it seems that the term really spans a continuum of possibilities.

Is it a DISPLAY grid ... or a display GRID?

DISPLAY grid to me implies that the primary usage/focus is around the concept of "display". Whether that is a web browser based environment to a single/few selected apps to a fully blown desktop environment. The focus is on the interactive usage. The grid portion than gets brought into the concept in two ways. First, the architecture used to deliver the display functionality could/may be based upon common grid architectures. Second is that there is a desire to use unused cpu cycles to do "tradional" grid computing tasks. This scenario would be typified by things like ... delivery of web based kiosks or "simple" web based vertical solutions.

display GRID than is the total opposite of the first scenario. The primary usage/focus is around the grid computing aspects and that secondary to the grid computing might be the display of the results of the grid computing tasks. This scenario would be typified by things like ... animation rendering, molecular modeling, dna sequencing etc... You might want to be able to display from the grid simply because the volume of the date is such that moving it would be tedious, or, it's a very iterative/interactive process that the user is involved in.

That's my thoughts... what do you think?

I finally started my "bandwidth monitoring" test from home. While I'm accumulating the statistics I thought people might be interested on how I'm gathering the stats.

Look back at my earlier blog about "Home networking gone wild" and you'll see a diagram of my home network. The goal of this test is to measure my inbound & outbound internet traffic and not the traffic that's local to the home network. The biggest "challenge" is that virtually all of todays "hubs" are really "switches". And the issue with switches is that you cannot view the traffic on all the ports from a single port.

So the solution is to track down an old ethernet hub. That is a real hub that does shared traffic across all the port. Fortunately a compatriot in my office had one stashed away and was willing to lend it to me for a little while.

With that accomplished I could now "instrument" my internet connection as follows.



With the Solaris 10 laptop plugged into the ethernet hub I can than run the open source tool called "ethereal" to snarf all the traffic between the internet and my home router.

I'm only a few days into gathering statistics when will report back what I find out.

Maybe you've heard Scott McNealy mention the idea of a "Display Grid" in relationship to both the current Sun Grid projects and Sun Ray / Tarantella.

For the past year now I've been running a project where we've placed a Sun Ray server infrastructure externally to Sun's corp network. With it we've been doing demos and setting up executive level contacts to experience first hand how Sun Rays work in a wide area network setting (WAN), using of course the worlds largest WAN... the internet. From some of this work we've been garnering ideas and use cases that may be used to develop some initial "Display Grid" ideas.

How well does this model work?

Here's a blog written by one of our test users you might find interesting.

I've had test users running Sun Rays from across the Atlantic with very good performance. The fact that Sun Rays run blazingly fast across a WAN has huge potentials in my opinion to change how companies deploy desktop computing. Even if you don't need to do it across the internet, think about the benefits within your own company network of being able to easily deploy and support desktops from a more centralized compute facility. (It's back to the future, mainframe computing, done with a much nicer, multimedia, more flexible display device).