New revenue opportunities for R&E networks and cyber-infrastructure

One of the growing challenges for many campuses around the world is how to accommodate the power and cooling requirements of cyber-infrastructure facilities such as high performance computers, storage facilities, etc.

Increasingly the costs of the bricks and mortar, power and cooling to house these facilities significantly outweigh costs of the actual cyber-infrastructure equipment.

In Canada, for example, many universities who are part of the HPC Consortium called Compute Canada will have to make significant investments in the coming year, to install and upgrade power and air conditioning systems to host a range of new computation facilities funded by CFI at various institutions.

The carbon emission impact has yet to be even taken into consideration in many any of these plans. The carbon footprint of a modern HPC facility can easily exceed the average use of several SUVs.

Global warming in not only a problem to be solved by politicians. It is a global issue in which we all have a personal responsibility to address regardless if we are an average joe citizen or world leading computer scientist.

Researchers and funding bodies need to take into consideration the carbon dioxide emission impact of all these cyber-infrastructure facilities. Building the fastest and best supercomputer regardless of its environmental impact is simply not an option any more. Universities and computing science researchers should be playing a leading role in identifying new cyber-infrastructure solutions which not only address their research requirements but also take into account the carbon emission impact of these facilities. Perhaps deploying energy efficient grids, sharing under-used computational facilities, or utilizing virtual computing is a better answer than building a physical cyber-infrastructure facility at every campus.

We also need to address the ongoing proliferation of computer clusters throughout various computer departments. Unfortunately most of these departments do not pay for the power and cooling costs associated with these facilities and so do not appreciate their true impact on the overall energy use of the university or the associated carbon emissions. As I mentioned on this blog before using Amazon's EC2/S3 service in many cases can be cheaper than the power costs alone of a modern computational cluster, never mind the operational and overhead costs of operating such a facility.

This is where regional and national research networks can play an important role. There are now many carbon offset companies who will audit programs that are designed to reduce carbon emissions. They will also broker payment of real dollars for the carbon reductions that result from the program. If an organization setups a tele-commuting program and demonstrate real and auditable reduction in carbon emissions they can earn revenue through the sale of carbon offsets to energy companies and other organizations. A good example is where IBM is working with a carbon offset company which is offering up to $1 million in carbon offsets for organizations to move away from their physical servers to high energy efficiency virtual servers operated by IBM.

R&E networks are ideally positioned to negotiate and implement these carbon offsetting solutions. Network organizations are essential for implementing any carbon offset strategy. As well the carbon impact of an optical R&E network is miniscule compared to the carbon footprint of many high performance computers and other facilities. The more we can use network bits and bandwidth for advanced science instead of physical facilities the greater the potential for earning valuable offset dollars (and I would argue the better the science community will be served).

Another potential carbon offset revenue opportunity is with distance learning and tele-medicine. Although the jury is still out on the pedagogical value of distance learning, encouraging students to undertake some of their course program work at home can be just as effective as tele-commuting in terms of earning carbon offset dollars. The same goes for tele-medicine. If companies can earn carbon offset dollars to implement tele-commuting programs, universities and R&E networks should be able to earn carbon offsets for offering distance learning and tele-medicine programs. (But as I argued in previous posts, rather than exchanging dollars in terms of carbon offsets, I would recommend exchanging other "zero carbon" awards such as offering participating students free eTextbooks, free music video, etc)

Finally R&E optical networks have an important role in redefining the entire value chain of the network itself. Many R&E networks are largely underutilized in terms of traditional measures of traffic volumes etc. Given these traffic volumes (and slowing growth) it would have been far cheaper in some cases for universities or funding agencies to purchase managed bandwidth from the carriers rather than build their own R&E networks.

But nobody yet has measured the carbon impact of these various optical, wavelength and customer owned networks. I would argue that in fact the carbon footprint of dark fiber, wavelengths and customer controlled network with optical switches is significantly less than a traditional carrier with expensive high end switches and (especially) routers which collectively consume the power of a small nuclear reactor. British Telecom for example has announced an initiative to use renewable energy sources as it is one of the biggest consumers of energy in the UK.

Instead of measuring the value of a network in terms of "bits per second", we instead should be using "bits per carbon". And while the utilization of R&E network may be low by traditional measurement standards of "bps" its impact on the environment may be significantly less when measured by "bpc" compared to a commercial network. And once again, the R&E networks can help develop a new business model through carbon offset trading by demonstrating that an optical lightpath mesh network has significantly less of a carbon footprint than a traditional electronic routed network.

An even more interesting and radical concept is to replace expensive electricity lines with optical networks. Instead of "wheeling" expensive power to physical servers at universities we can instead "wheel" inexpensive bits between virtual servers, grids located at renewable energy sites around the world.

For example the global community of optical research networks (GLIF) could build a "follow the sun" grid infrastructure. Solar powered high performance computing facilities could be located at remote desert locations throughout the world. But these systems would not be connected to any electrical grid, and instead be linked by a global high speed optical network. As the sun starts to set on any given HPC site, the currently running jobs and OS images would immediately transferred over the optical network to the next HPC site that is just starting to come active with the rising sun.

Bottom line is that I believe research and education networks can play important leadership role in defining these new business and network models related to trading "bits and bandwidth for carbon". They could also be working with universities to freeze, if not decrease, the carbon output of these institutions. To my mind universities should at the forefront in our society in finding solutions and new business models to address global warming. At least they should not be the worst offenders in terms of all these high carbon emission cyber-infrastructure facilities that are now being deployed at our campuses.