Truly Advanced Computing – A Visit to TACC

One of the real highlights of the HPC Consortium was a visit to the Texas Advanced Computing Center (TACC). TACC is part of an extended US large computing facility group known as the TeraGrid. The TeraGrid includes many of the university-related supercomputing facilities (including the Pittsburgh Supercomputing Center, which is near-and-dear to my historical educational heart) as well as the national labs (Oak Ridge, Los Alamos, and so on). This extended grid enables researches to share information and projects across the country. At the core of TACC is a system known as Ranger. You can read more about it here.

From my perspective, two things are really important about the visit to TACC:

1. When you see a truly big computing center, it changes your perspective on what you do.
2. Back when I first started working on computers, a computer of approximately the same physical size as the whole Ranger system had about as much computing power as an iPod has now (or even your cell phone).

Perspective

When you visit the Ranger room, you see a lot of hardware and a lot of green. I’ve known about the “sea of green” concept for years, but to see it on the scale of Ranger is just amazing. Literally, at a glance, you can see the status of any hardware component in the room. If you aren’t in the room, of course, you use remote software monitoring tools to show you the virtual representation of the same thing.

My brain working the way it does, I then ask “Well, could you do this in Second Life (or some similar virtual world)?” In other words, you build a representation of your physical data center in the virtual space (using, I assume, a simple click-and-drag interface). This virtual representation communicates via the appropriate management protocols to get status information from your real-world data center. Then, in the virtual world (when you aren’t physically in the data center), you can “walk through” your data center and see the status at a glance. It would be cool. Would it be better than some of the current, less graphic-intensive, less realistic representations of the same concept? I don’t know. What do you think?

The other interesting thing that came to my mind was what this kind of environment means to information. The traditional approach to information products has been to provide detailed documentation about individual components in the system (hardware and software). It is then largely left up to the user to figure out how to put those components together in their own environment and to then write down anything that they think is important. To a certain extent, we will never change this approach (and shouldn’t). However, the question that I have to ask is whether users would be better served if we spent time on documentation some of the “best practices” around how to put components together into a larger solution and how the customer can use that to achieve certain real-world/job tasks.

Computing Power

Enough said. ☺ The first time that I worked on a computer, it was an old DEC 10 system. A few years later, when I started working at DEC, VAXes had come out, and they had a lot more computing power for the physical space. When you then see a room the size of the Ranger room at TACC, and calculate the amount of computing power now in that space, the mind is filled with the amazing leaps that we have made in technology over the past 30 years. A couple of the presenters at the Consortium asked “Where will we be in 10 years?” That is an interesting question, and a topic for several more blogs.

Posted at 01:36PM Nov 18, 2008 by jcinfoblog in General  |  Comments[0]

Everybody's Workin' on the Weekend....

Taking the words from a 1980s Loverboy song and changing them slightly, a large number of Sun employees, partners, customers, and students are "working" this weekend at the HPC Consortium in Austin, TX. There's so much to talk about that it's going to be difficult for me to decide what to mention. Of course, I'm not the only person who's blogging this event, and some folks are on twitter, too.

Before I start my various blogs about the actual Consortium and the content of the presentations, though, I should probably talk about some changes in my job over the past couple of months which would help you to understand why I'm at this Consortium. As I had mentioned in a previous blog entry, Sun Grid Engine 6.2 had reached a significant milestone. So, once that milestone was reached, it was time for me to go apply my skills to some other products. One of my current projects keeps me closely affiliated with Sun Grid Engine, while the other is in a different area.

The first project is really not a new project, but it has advanced into some different areas. Last Spring, I became involved with a small project to provide a preview stack for people who want to develop HPC applications on Solaris. Sun Grid Engine is one of the components that comprise this stack. For information about the Sun HPC Software for Solaris project, see the HPC developer product community on OpenSolaris and the associated information wiki.

The other project on which I am working is an update to the Sun SAM-QFS software. As some of you know, I have spent quite a bit of time in various storage software projects during my career (some in a "previous life" at a company once known as DEC). So, it's nice to return back to the storage software world. I can't say a lot about that project yet, but from an information products' perspective, watch this space.

So, now that you know what I've been doing, it might be obvious why I am attending this particular conference. The link between HPC and SAM-QFS might not be as obvious as the link for the HPC Software for Solaris. However, the various features of the SAM archive management product and the QFS file system are used in many existing HPC environments. Furthermore, future features (such as object support on OpenSolaris) might make these products even more effective for HPC purposes.

Okay, that's enough to get this group of entries started. I'll have more to add over the next couple of days. Stay tuned.

Posted at 11:02PM Nov 15, 2008 by jcinfoblog in General  |  Comments[0]