måndag 24 februari 2014

Networking in the Long Emergency

I met Barath Raghavan a few days ago when he visited UC Irvine and gave a talk about his RiSCIT-related research interests. I really appreciated listening to Barath and to (finally) meet him in person.

I don't know how it came about that I originally stumbled upon Barath's paper "Networking in the long emergency" (pdf file here). I think I must have found his paper when I searched for research papers in the intersection of "peak oil" and "computers" [something]. The title immediately told me this was a paper I definitely should read as it obviously must have been written by someone who was familiar with Kunstler's book "The long emergency" (2006). That was absolutely right and the paper didn't disappoint me. In fact, I thought is was very good - it was obviously written by someone who had thought long and hard about these issues.

I read the paper in April in 2012 and after having finished reading it immediately sent a mail, reaching out to Barath. Some time later I suggested we should think about writing something together, but the timing was all wrong as Barath had just left the academy for work on data center networking at Google. This is however how Barath presented himself in a distribution list I (do think I) recommended him to join in September 2012:

"I'm Barath Raghavan and I've been on the networking, systems, and security side of computer science but have had a parallel side interest in environmental science and energy since I was an undergrad at UC Berkeley. Learning about peak oil around 2007 made med delve deeper into how the various problems facing the globe today converge, and has really been a full-time passion since then.

I finished my PhD from UCSD in 2009, and went to Williams College to teach for a year as a visiting professor. While I was there, I decided to try to convey what I had been learning to the undergrad CS students at our weekly seminar in a talk I called "Computing in the Long Emergency". [...] my talk [...] led my colleague Justin Ma and I to write a paper, "Networking in the Long Emergency" that focused more on networked systems research, since that's the community we're part of.
I've recently joined Google (as has Justin), but I hope to continue pursuing this line of thinking and discussion with you all"

Things petered out between us after that and we haven't been in contact for 18 months or so (besides being part of the same e-mail distribution list) and I know from personal experience how academic pursuits can turn into a very distant concern when you are working in industry. As it turns out, Barath recently left Google (after 18 months there) and now works 50% in the academy (at UC Berkeley, at ICSI - the International Computer Science Institute) and 50% in a nonprofit R&D organisation he started.

In his UCI talk last week, Barath described his 2011 "Networking in the Long Emergency" paper (pdf file here) as the starting point for his explorations in the area. Let's call that paper the FOUNDATION as it has since been followed by four more paper:

- "The energy and the emergy of the Internet" (2011, pdf file here). Here's what I wrote about this article in this blog: "Very interesting attempt to estimate "how much energy is required to construct, run, and maintain the Internet". Includes both "running costs" as well as the emergy - the energy that is "embodied" in the Internet's constituent parts (i.e. the energy needed to construct cell towers, routers, end devices etc.)." This is the follow-up ANALYSIS paper.
- "Macroscopically sustainable networking: An Internet Quine" (2012, pdf file here). I read this paper last spring but haven't come around to writing about it (yet) on the blog. The paper asks how we could built a future low-tech, low-cost "minimalistic" Internet if the need should arise. This is the first of two follow-up DESIGN papers. Our CHI 2014 sustainability workshop position paper ("Usability as a threat to a sustainable future") is partially derived from ideas in this paper.
- "An intermittent Internet architecture" (2012, pdf file here). Here's what I wrote about this article in this blog: "Very interesting thought-experiment to "re-design the Internet for an energy-constrained future powered by diffuse, intermittent, and expensive power sources." This is the second of two follow-up DESIGN papers.
- "Networking for undeveloping regions" (2013) was written for an academic audience but was later (April 2013) posted on Barath's blog, Contraposition. This is the follow-up POLEMIC paper. Do note that the paper refers to "undeveloping" (not "underdeveloped" or "developing") regions and that there's a big difference between these terms, despite referring to the same geographic regions...

I should at this point also mention that Barath's blog, Contrapositon, is a good read. Barath writes it together with a friend, but it's mostly Barath who writes stuff. The current volume is regular but very low, with only slightly more than one new blog post per month last year. It's however possible to see a stark reduction of blog posts at the time when Barath started to work for Google, so perhaps the pace will pick up now that he has switched back to the academy...?

I think there are large overlaps between Barath's interests and perspectives and my own. We even seem to hold the same author, John Michael Greer, in the highest regard. An underlaying question that we are both interested in is how to reconcile 50 years of exponential growth in computing (Moore's law and all that) with the bell curves that characterise limits to growth, depletion of natural resources and limits/scarcity. What will happen when runaway developments in a field (computer science) start to bump up to irreversible natural (for example energy) limits?

In short, I very much recommend all the five texts above. Both me and Barath are furthermore interested in deeper issues than just "how can we make routers [or some other small and separate part of the larger network] more energy efficient?", or, "how can we make this smartphone more energy-efficient so that it runs longer per charge?" (...and so that we can sell twice as many new phones). Not all green computing is green in any deeper sense of the word and some (much?) could be characterised as greenwashing or worse - if we only had better metrics for determining what actually constitutes "green" in a less narrow sense of the word.

Barath currently spends half his time in a nonprofit organisation he created, De Novo Group. It looks a lot larger that it is if you check out the website, but they are (as far as I understand) basically funded by Google to do applied work in figuring out and building/testing solutions for wireless, limited-bandwidth networking in developing regions such as rural parts of Asia and Africa. This work is partially in line with Barath's interest in the (lack of) resilience of our current computing infrastructure. The Internet is nowadays very dependent on a handful of very large key companies to run smoothly. These companies are very good at what they are doing, but they are not failsafe. Having one big, fat Internet for the whole planet makes it very vulnerable, or rather, brittle. If it for some reason would fail, it fails utterly and catastrophically. The problem is that resilience equels "redundancy" and "inefficiency" and thus often comes into conflict with managing (trimming) costs. Why pay extra when, on a day-to-day basis, something works "almost always"? Yes, indeed, why would it make sense to build a more resilient system? Perhaps because the effects of that one indeed very unlikely event are so catastrophic that it would be worth anything to avoid it (but how do you make that case beforehand?). Resilience could thus imply several independent, perhaps loosely connected Internets (plural) - which is just the thing Barath is working on in De Novo.

It is, unfortunately, easy (or at least easier) to get funding for "the next big thing" than to get funding for thinking about a Plan B (if "the shit hits the fan"). A "developing regions" project is however very similar (or the same) as an "undeveloping regions" project and who know, perhaps the "counter-ICT4D movement" that I speculated about in my previous blog post will come true sooner than we think?

Inga kommentarer:

Skicka en kommentar