Since 1986 - Covering the Fastest Computers in the World and the People Who Run Them

Language Flags
November 3, 2006

How to Talk to a Techno-Liberal (and you must)

Michael Feldman

With election day right around the corner, I started to think about what it means to be a conservative versus a liberal and how that impacts one's world view. At the most basic level, conservatives want to support continuity. In general they're happy with the status quo and want to maintain it. Conservatives act as society's buffer for a changing world. Liberals, on the other hand, embrace change. By nature, they're not satisfied with the way things are and want to find new ways to do things. Liberals are the opportunists that move a society forward.

Before I go any further, I have to admit that I'm actually equating liberalism with progressivism — a popular notion, but technically inaccurate. Liberalism is actually more about civil liberties and individual rights. But for the purposes of this discussion we're talking about liberal behavior, not ideology.

In fact, while we normally associate liberalism and conservatism with politics, they really represent socio-cultural perspectives that map to all human endeavors. This includes how we develop and use technology, especially computer technology. I'll get to this shortly. But first let's look at the demographics.

According to the Pew Research Center, in their latest poll among registered voters, conservatives outnumber liberals by a two-to-one margin — 38 percent conservative versus 19 percent liberal. Another 38 percent identify themselves as moderates. By party, 32 percent of Democrats describe themselves as liberals, 23 percent describe themselves as conservatives and 41 percent categorized themselves as moderates. The respective numbers for Republicans are 5 percent, 64 percent and 29 percent.

One of the unfortunate effects of the concentration of liberals in the Democratic party and conservatives in the Republican party is the open warfare that now exists across this philosophical divide. In this culture of two-party tribalism, each party strives to annihilate the other. I say unfortunate because it's detrimental to have a political system dominated by either liberalism or conservatism. As can be deduced from my opening paragraph, I believe both philosophies have important roles to play in politics.

And not just in politics, but in technology as well. I've come to see three basic philosophical approaches to science and engineering, reflected by these common utterances:

  1. “If it ain't broke, don't fix it.” (techno-conservative)
  2. “Let me tweak it a little bit.” (techno-moderate)
  3. “We need a whole new paradigm.” (techno-liberal)

In the technology arena, liberalism maps to innovation and risk-taking; conservatism to standardization and risk avoidance. Without innovation, progress stops. But too much causes chaos and confusion (think the 60s). On the other hand, without standardization and technological continuity, getting anything done becomes impractical. But too much limits choice and kills innovation (think the 50s). Obviously, what we need is some sort of balance.

Do we have balance now? Depends on who you are.

If you're a techo-conservative in IT, everything is moving too fast. Server OEMs are in constant motion trying to keep pace with the latest processor and OS releases. On the software side, ISVs struggle to keep up as new hardware platforms are developed and older ones die off. In the 1980s, Digital Equipment Corporation (DEC), at the time, the second largest computer company in the world, was swept away when its proprietary VAX/VMS and PDP-11/RSX-11M systems became obsolete after the rise of standard Unix platforms and commodity PCs. Today, Microsoft is trying to reinvent itself as the desktop platform becomes subordinate to the Internet platform. Yes, being a techno-conservative has its downsides.

On the other hand, if you're a techno-liberal in IT, the future can't come soon enough. For those who see the true promise of the Web for multi-media and as a general platform for application software, the Internet is still far too slow and primitive. If you're a neurobioloist who wants to simulate the human brain at the molecular level, it's going to be pretty frustrating until you get the right software tools and the petaflops to back them up. Beyond that, Ray Kurweil, the ultimate tech progressive, envisions molecular computing and spiritual machines to provide a complete redefinition of the human condition. If being a techno-conservative is stressful, being a techno-liberal is discouraging.

You might think that the high-tech crowd, especially the IT industry, would be dominated by liberal-thinkers, since innovation is the foundation of engineering. But that's probably only true for areas that haven't been commercialized yet, like nano-engineering. Today, most computer technology is so well integrated into the commercial realm that the conservative tendencies of businesses drive information and computer technology. Earlier this year in our sister publication, GRIDtoday, Tom Gibbs observed that IT invests around 15 percent in innovation. He thought that was dangerously low, but it probably reflects the natural conservatism of commercial enterprises. And even at this level of investment, IT has delivered enormous value to the economy and has been the driving force behind rising productivity for decades. For better and for worse, commercial IT buffers the rate of innovation.

In our world — high performance computing — it's easy to see the philosophical divisions that define the community. Toward the conservative end of the spectrum we have commodity clusters, the x86 ISA, Ethernet, Linux and the reams of code written in Fortran, C and MPI. On the liberal end we have multi-core processors, hardware accelerators (FPGAs, FP coprocessors, GPUs), InfiniBand, new parallel programming languages, national petascale programs, heterogeneous computing and HPC research.

At last week's HPC User Forum in Manchester, the tension between techno-conservatism and techno-liberalism was in evidence. Just some examples:

  • IDC's Addison Snell reported that about one-third of HPC users are looking at accelerator processors, mainly FPGAs.

  • Numerical Algorithms Group's Ian Reid said the days of the single-core treadmill are over, and multi-core is here to stay. He said that this creates major software issues that will force us to move to hybrid hardware architectures. But software portability must be maintained. He also noted there is considerable concern about whether the HPCS languages will deliver on their promises.
  • Intel's Stephen Wheat talked about the return of hyper-threading, 80-core teraflop chips and silicon photonics, but lamented that the impending petaflop hardware will arrive before the software is ready.
  • Paul Muzio, AHPCRC/NCSI, stressed that GM, Dassault and many other major companies are still using Fortran. He maintained that businesses won't throw out those huge investments. The applications proposed for petascale computing are not the ones companies or the defense establishment will invest in. But there are opportunities for languages like Fortran to evolve, such as Co-Array Fortran.
  • Andy Grant said that IBM is starting to see requirements for accelerators in procurements. IBM is installing a large Opteron cluster with ClearSpeed boards at the University of Bristol. He also talked about the Blue Gene/L successors — Blue Gene/P (for petaflop), followed by Blue Gene/Q (10 petaflops).
  • Manchester's Andrew Jones chaired a panel on whether programming model changes are needed for petaflops computing. Jones noted that scaling to 1,000 processors on homogeneous architectures is difficult today. Petascale and exascale computing will involve many more threads than today, and possibly heterogeneous architectures. He believes we may need a new programming paradigm.

It's actually gratifying to see this tension in HPC. The real danger would be to descend into techno-Democrats and techno-Republicans. As long as the conservatives don't slow innovation too much and the liberals don't send us into chaos, the community will move forward.

I'm Michael Feldman and I approve this message.


As always, comments about HPCwire are welcomed and encouraged. Write to me, Michael Feldman, at