Reed's Ruminations: A Blog by Dan Reed

Broadband: Oxygen for a Digital World

Much has been written about what constitutes the baseline for acceptable broadband network speeds. The U.S. National Broadband Plan includes the following recommendations:

Every household and business location in America should have access to affordable broadband service with the following characteristics:

• Actual download speeds of at least 4 Mbps and actual upload speeds of at least 1 Mbps
• An acceptable quality of service for the most common interactive applications

The FCC should review and reset this target every four years.

The latter – reviewing and updating the target regularly – is essential. What was broadband in the past is woefully inadequate today. Not long ago, a 1200-baud dial modem was an object of technological envy, and a 64 Kbps ISDN connection was beyond the reach of mere mortals. Some of us even remember acoustic couplers. Today, they are all artifacts of our Internet history.

Supercomputing and Internet History

Today's Internet traces its origins to the early ARPANET and its progeny, notably NSFNET. Although both the ARPANET and the early NSFNET were continent spanning networks, the ARPANET backbone was based on 56 Kbps links. That remained the standard backbone speed even as NSFNET was deployed to interconnect the U.S. supercomputing centers.

I was at the University of Illinois when we began the NSFNET transition to a T1 (1.5 Mbps) backbone at NCSA. The iconic image of the early NSFNET T1 backbone, shown here, was the work of my former NCSA colleagues Donna Cox and Bob Patterson. The world changed quickly after that, with deployment of a new T3 (45 Mbps) NSFNET backbone, birth of the Mosaic web browser fueling the dot.com boom and rapid deployment of dark fiber.

By 2001, Charlie Catlett and I were negotiating access to 4 10 Gbps lambdas for the NSF TeraGrid, connecting NCSA in Illinois and SDSC in California. We had initially pursued options to light ten lambdas (i.e., 100 Gbps) because we were worried that 40 Gbps might not be enough to support large-scale data sharing and high fidelity scientific visualization. We were both right and wrong. The scientific data did grow, but the scientific applications took longer to develop.

At about the same time, when I picked the name TeraGrid for the distributed grid we were building to connect the NSF supercomputing centers, Ian Foster asked me if we might later regret using the "tera" prefix in a world were supercomputing performance changed so rapidly. I replied that I thought it would last long enough, and is has, far longer than I might have expected.

Though we are now deep into the petascale computing era, we are by no means in the terabit era. Although network backbone speeds have continued rise, data volumes and user expectations have risen even more rapidly, both in scientific computing and in business and consumer web services. We are now building petabyte islands and content distribution networks (CDNs) connected by relatively slow multi-gigabit networks. I keep wondering when multi-terabit networking will be commonplace.

A Devices and Services World

Inexpensive broadband access has made possible Internet searches, cloud services and always-connected PCs, tablets and smartphones. It has transformed our society, shifted our notions of commerce, shaken governments, and brought McLuhan's global village within reach.

Thus, it is no exaggeration to say that broadband networks, wired and wireless, are the oxygen that lets this world of devices and services breathe. However, the air remains perilously thin. Billions of devices are being added to the network, data volumes are rising exponentially, and expectations for anytime, anywhere access are growing worldwide. For many, digital inclusion remains a dream, rather than a reality.

If we are to meet burgeoning demand, extend the benefits of broadband services to all of the world's population, and enable a new generation of data-rich services and experiences, we need to approach the problem of broadband deployment in new ways.

We need both top down and bottom up strategies that encourage innovation and more rapid deployment of last mile connectivity. We must rethink our approaches to spectrum allocation and access, embracing more nimble and adaptive schemes. We must expand the capabilities of our backbone networks to meet demand.

Above all, we need to remember that the definition of broadband is ever shifting. We will always want more.

Analog Computing: Time for a Comeback?

N.B. I also write for the Communications of the ACM (CACM). The following essay recently appeared on the CACM blog.

In the early dates of the automobile, there was a lively competition among disparate technologies for hegemony as the motive power source. Steam engines were common, given their history in manufacturing and locomotives, and electric vehicles trundled through the streets of many cities. The supremacy of the internal combustion engine as the de facto power source was by no means an early certainty. Yet it triumphed due to a combination of range, reliability, cost and safety, relegating other technologies to historical curiosities.

Thus, it is ironic that we are now assiduously re-exploring several of these same alternative power sources to reduce carbon emissions and dependence on dwindling global petroleum reserves. Today's hybrid and electric vehicles embody 21^st century versions of some very old ideas.

There are certain parallels to the phylogenic recapitulation of the automobile now occurring in computing. Perhaps it is time to revisit some old ideas.

Analog History

Use of the word "computer" conjures certain images and brings certain assumptions. One of them, so deeply ingrained that we rarely question it, is that computing is digital and electronic. Yet there was a time not so long ago when those adjectives were neither readily assumed nor implied when discussing computing, just as the internal combustion engine was not de rigueur in automobile design.

The alternative to digital computing – analog computing – has a long and illustrious history. Its antecedents lie in every mechanical device built to solve some problem in a repeatable way, from the sundial to the astrolabe. Without doubt, analog computing found its apotheosis in the slide rule, which dominated science and engineering calculations for multiple centuries, coexisting and thriving alongside the late comer, digital computing.

The attraction of analog computing has always been its ability to accommodate uncertainty and continuity. As Cantor showed, the real numbers are non-countably infinite, and their discretization in a floating-point representation is fraught with difficulty. Because of this, the IEEE floating-point standard is a delicate and ingenious balance between range and precision.

All experimental measurements have uncertainty, and quantifying that uncertainty and its propagation in digital computing models is part of the rich history of numerical analysis. Forward error propagation models, condition numbers, and stiffness are all attributes of this uncertainty and continuity.

Hybrid Futures

I raise the issue of analog computing because we face some deep and substantive challenges in wringing more performance from sequential execution and the von Neumann architecture model of digital computing. Multicore architectures, limits on chip power, near threshold voltage computation, functional heterogeneity and the rise of dark silicon are forcing us to confront fundamental design questions. Might analog computing and sub-threshold computing bring some new design flexibility and optimization opportunities?

We face an equally daunting set of challenges in scientific and technical computing at very large scale. For exascale computing, reliability, resilience, numerical stability and confidence can be problematic when input uncertainties can propagate, and single and multiple bit upsets can disturb numerical representations. How can we best assess the stability and error ranges on exascale computations? Could analog computing play a role?

Please note that I am not advocating a return to slide rules or pneumatic computing systems. Rather, I am suggesting that we step back and remember that the evolution of technologies brings new opportunities to revisit old assumptions. Hybrid computing may be one possible way to address the challenges we face on the intersecting frontiers of device physics, computer architecture and software.

A brave new world is aborning. Might there be a hybrid computer in your hybrid vehicle?

Surrounded By Opportunities

Perhaps the closest I have approached the feeling of intellectual omnipotence (and it has not been often) was when I was a graduate student, teaching undergraduate FORTRAN programming for science and engineering students. Those were the days of punched cards, batch processing and line printer output. It was a time when dinosaurs roamed the earth, my slide rule was still a fond memory, and we were computing on massive machines with less performance than my smartphone has today. (Ignore that part about the dinosaurs, as it was a wee bit earlier.)

During my office hours, students would troop through for help with their programming assignments, bringing program printouts. These printouts began with a "day file" showing the timeline of compilation and execution steps (JCL steps on IBM S/360), followed by the program and then any output. After having seen a few students, you realize that elementary programming errors have just a few equivalence classes, often identifiable from the day file alone. I took great pride in listening to the student description of the problem, predicting the root cause and then immediately locating it on the printout. The student was inevitably amazed. Yes, it was a cheap parlor trick, but as a graduate student, you take gratification where you can find it.

All of this reminds me that sequential programming remains difficult and parallel programming even more difficult, despite all our efforts to tame the unruly beast with syntactic sugar, specification and scripting languages, libraries and tools. I am no expert in parallel programming language design, nor am I a cognitive psychologist, but like many of my readers, I have pulled a few all-nighters writing code. Perhaps that at least qualifies me to have an opinion.

In a world of parallel programming, multicore architectures, dark silicon, and rising software complexity, we face some difficult choices. It's time to learn from the past.

Fading Abstractions

We have been fortunate to date. Although sequential programming at scale is intellectually challenging, we have been riding the wave of ever-increasing processor performance, made possible by semiconductor advances. In turn, these were based on an abstraction – the Mead-Conway design rules –that largely separated device physics from microprocessor architecture. Finally, the von Neumann abstraction has also served us well, allowing decades of software development atop an easily understood model of sequential instruction execution.

Our run of good luck is over. Put another way, we are surrounded by opportunities! Parallelism is now the norm for almost all software developers, not just those of us in high-performance computing. More perniciously, heterogeneous parallelism is here, with disparate cores, accelerators and functional units. To manage dark silicon and exploit continuing transistor density increases, this trend will only accelerate, with a dizzying array of heterogeneous chips and devices challenging code portability and performance tuning. Finally and most worrisome, our device physics abstractions are breaking down, with device physics affecting architecture. We face real silicon reliability, process variation and energy challenges, and near threshold voltage computations emerging.

Holistic Collaboration

The abstractions of device physics, processor architecture and programming models naturally led to a corresponding intellectual specialization. Consequently, those working in materials science and device physics have little in common with those designing software tools. I believe that must change if we are to address our challenges.

We need new, integrated design methodologies and rich collaborations that consider end-to-end design and capability. In many ways, this is a return to the past, necessitated by the desire for continuing increases in performance, mediated by the desire for reliability, scalability, programmability and economic feasibility. It will not be easy, but it is a way forward.

It Is Fall: There Must Be Clouds

Today, I testified to the Subcommittee on Technology and Innovation of the U.S. House of Representatives Committee on Space, Science and Technology. The topic of the hearing was "The Next IT Revolution?: Cloud Computing Opportunities and Challenges." At the hearing, I was joined by Mike Capellas (CEO, VCE), Nick Combs (Federal CTO, EMC), and David McClure (Associate Administrator, GSA).

This hearing drew on some of the content from the TechAmerica Commission on the Leadership Opportunity in U.S. Deployment of the Cloud (CLOUD²), on which I had the privilege to serve as the commercial vice-chair. In addition, my own remarks were augmented by Microsoft's cloud engagement experiences with the U.S. National Science Foundation and other research funding agencies. (See Innovation Via Client Plus Cloud: Microsoft-NSF Partnership for background.)

My complete written testimony will ultimately appear on the Subcommittee web site here, and the hearing charter, the subcommittee press release, and the hearing video capture most of context. However, there are a few points from my testimony that seem appropriate to repeat here. These concern research empowerment via the cloud, support for basic research and education and broadband expansion.

Accelerating Scientific Discovery for Research via the Cloud

Throughout the history of science, data has been scarce and precious. Indeed, the modern scientific method is defined by a careful cycle of hypothesis and experiment, which gathers experimental data to test the hypothesis. In a few short years, scientists and engineers have gone from scarcity to an incredible richness, necessitating a significant change in how they manage and extract insight from all this data. In a parallel shift, many of our scientific, engineering and societal questions increasingly lie at the intersections of traditional disciplines.

Increasing data volumes and the complexity of collaboration on interdisciplinary problems are challenging our historical approaches to discovery and innovation via computing. Most researchers and research institutions are ill-prepared for the large-scale computing infrastructure management challenges posed by large data sets and complex models. The cloud and associated applications and tools offer a possible solution to this challenge by letting scientists be scientists.

The U.S. government can accelerate this transition by encouraging the purchase of cloud services, in addition to the acquisition of local IT infrastructure, and by supporting new tools that facilitate distributed collaboration and simplify access to multidisciplinary scientific data. As I have noted before, Microsoft is acting on this belief, working in partnership with the National Science Foundation.

Fostering Continued Support for Computing Research and Education

Today's cloud technology is derived from basic computing research conducted over the past four decades. To ensure that the U.S. continues to remain at the forefront of cloud technology, continued investment in basic research is critical. There are deep and open questions in areas as diverse as the future of silicon scaling and system-on-a-chip design, energy-efficient systems, primary and secondary storage, data mining and analytics, wired and wireless networks, system resilience and reliability, privacy and security, and user interfaces and accessibility, to name just a few. Insights and innovations from this research will spawn new companies, create jobs and reshape our future.

In addition to continued research investment, it is critical to support the pipeline that produces researchers, and others who will able to invent new uses of the cloud and information technology. The U.S. Bureau of Labor Statistics estimates that the computing sector will have 1.5 million job openings over the next 10 years, yet the number of graduates receiving Bachelors, Masters or Ph.D. computer science degrees is far short of that. In addition, we must strengthen the quality of and access to computing education at all levels. Consistent with these concerns about the IT workforce and computing education, Microsoft is a founding member of the Computing in the Core coalition, which supports computer science education, particularly at the K-12 level.

Broadband: The Need for Speed

The web and cloud services depend on broadband communications. Without them, service and information sharing are impossible. Concomitantly, ensuring reliable wired and wireless connectivity, with adequate bandwidth and latency, is critical to ensuring successful adoption of the cloud and realization of its benefits. The phenomenal growth of digital data, the rise of streaming media services, and the explosive growth of Internet-connected devices are all straining our nation's broadband infrastructure.

It is critical that we continue to design and deploy new backbone networks that support higher data rates, develop and deploy new protocols and infrastructure for the next generation of wireless networks and define the global standards that will shape the future of the globe-encircling cloud. We must also remember that digital access to information and services is increasingly the enabler of economic competitiveness, of lifelong education in a rapidly changing world, and of government efficiency and service delivery.

21^st Century Innovation

This is an exciting time, when the future becomes the present. Computing can be a great equalizer, providing access to the world's knowledge base to individuals, anywhere, anytime; empowering entrepreneurs and companies large and small to sell their products and ideas globally; and enabling scientists and engineers to discover and innovate in ways that will define the future.

Why We Compute

N.B. I also write for the Communications of the ACM (CACM). The following essay recently appeared on the CACM blog.

Why do we, as researchers and practitioners, have this deep and abiding love of computing? Why do we compute? Superficially, the question seems as innocuous as asking why the sky is blue or the grass is green. However, like both of those childhood questions, the simplicity belies the subtlety beneath. Just ask someone about Raleigh scattering or the quantum efficiency of photosynthesis if you doubt that simple questions can unearth complexity.

At its most basic, computing is simply automated symbol manipulation. Indeed, the abstract Turing machine does nothing more than manipulates symbols on a strip of tape using a table of rules. More deceptively, the rules seem simpler than some board games. Though vacuously true, the description misses the point that symbol manipulation under those rules captures what we now call the Church-Turing thesis.

However, as deep and as beautiful as the notion of computability really is, I doubt that is the only reason most of us are so endlessly fascinated by this malleable thing we call computing. Rather, I suspect it is a deeper, more primal yearning, one that underlies all of science and engineering and that unites us in a common cause. It is the insatiable desire to know and understand.

Lessons from Astronomy

When I recently I stood atop Mauna Kea, looking at the array of telescopes perched there, I was again struck by our innate curiosity. Operated by a diverse array of international partnerships and built on Mauna Kea at great expense, they are there because we care about some fundamental questions. What is the evolutionary history and future of the universe? What are dark matter and dark energy? Why is there anything at all?

Answers to these questions are not likely to address our current economic woes, improve health care or address our environmental challenges. We care about the answers, nevertheless.

As I pondered the gloaming sky, my thoughts turned to Edwin Hubble, who first showed that some of those faint smudges in the sky were "island universes" – galaxies like our own. The universe was a far bigger place than we had heretofore imagined. As Hubble observed about this age-old quest to understand:

From our home on the Earth, we look out into the distances and strive to imagine the sort of world into which we are born. Today we have reached far out into Space. Our immediate neighborhood we know rather intimately. But with increasing distance our knowledge fades, and fades rapidly, until at the last dim horizon we search among ghostly errors of observations for landmarks that are scarcely more substantial. The search will continue. The urge is older than history. It is not satisfied and it will not be suppressed.

Hubble's comment was about the observational difficulties of distance estimation and the challenges associated with identifying standard candles. However, it could just as easily have been a meditation on computing, for we are driven by our own insatiable desires for better algorithms, more flexible and reliable software, new sensors and data analytics tools, and by ever larger and more faster computers.

Computing the Future

Why do we compute? I suspect it is for at least two, related reasons, neither relating to publication counts, tenure, wealth or fame. The first is the ability to give life to the algorithmic instantiation of an idea, to see it dance and move across our displays and devices. We have all felt the exhilaration when the idea takes shape in code, then begins to execute, sometimes surprising us in its unexpected behavior and complexity. Computing's analog of deus ex machina brings psychic satisfaction.

The second reason is that computing is an intellectual amplifier, extending our nominal reach and abilities. I discussed the power of computing to enable and enhance exploration in another CACM blog. (See Intellectual Amplification via Computing.) It is why those of us in computational science continually seek better algorithms and faster computer systems. From terascale to petascale and the global race to exascale, it is a quest for greater fidelity, higher resolution and finer time scales. The same deep yearning drives astronomers to seek higher resolution detectors and larger telescope apertures. We are all chasing searching the ghostly signals for landmarks.

It is our ability to apply our ideas and their embodiment in code to a dizzying array of problems, from the prosaic to the profound, that attracts and compels us. It is why we compute.

Hubble was right. We compute because we want to know and understand. The urge is deep and unsatisfied. It cannot be denied.