Architecture in the Cloud
What's at stake for architects in the emergence of new computing platforms
by Mark Collins
This article originaly appeared in CC:, a publication of the Columbia University GSAPP. You can find the article here.
Type “Architecture” into Google and press the “I’m Feeling Lucky” button. Predictably, you are taken to Wikipedia’s article where you will be enlightened about the “Post-modern design of Santiago Calatrava” and regaled with a closing reference to the fictitious “Dynamic Tower” as an exemplary example of sustainable architecture.1
Factual inconsistencies and lack of definitiveness are not unique to the architecture page on Wikipedia, but it hints at the lack of debate, or even presence, of architects in the emerging cloud of information. If we aren't tending to our informational backyard, what are we doing in the cloud? Does the cloud really matter to architects?
A start would be to look at how cloud computing will transform the design process. One benefit that is actively being realized right now is a significantly reduced wait for those beautiful, photo-realistic renderings that were ubiquitous in the last bubble economy. The NVIDIA Reality Server2
is one of many cloud architectures that are attempting to deliver real-time rendering, scalable to vast numbers of resources and accessible even through lightweight platforms such as phones and tablet computers.3
This particular architecture calls for building data to be stored on a remote server, which streams visual content back to each client. The Reality Server is an ambitious but appropriate title when one considers the fine grain exploration of even city-scaled urban environments that an early demo shows. The reality server isn't so much interested in reality as such but, rather, projecting a newly designed reality and letting users explore even its most banal interiors and alleyways. The unlimited aspect of cloud rendering technology allows even someone viewing on a smart phone to be treated to physically accurate lighting, dynamic shadows, and real-time animation.
"If a problem can be decomposed into millions of smaller problems with little interdependency, then it can easily be farmed out to the cloud. Many computational problems in architecture can be formulated to operate within this framework."
Raytracing is one of a host of processes that can be effectively distributed - or solved - in parallel. So called “share nothing” architectures are based on the atomization of both thought and action. If a problem can be decomposed into millions of smaller problems with little interdependency, then it can easily be farmed out to the cloud. Many computational problems in architecture can be formulated to operate within this framework; for instance, in fluid dynamic computations that model the air flow patterns around a skyscraper, or finite element analysis that can calculate the stress intensity and distribution in a loaded structure. These algorithms work through decomposing objects into tiny voxels or triangles while considering just the local forces at work at each point. Clouds are especially adept at solving these kinds of problems and often make it a requirement that is a consequence of how they architect their interfaces. This is a problem solving paradigm of silo-ed workers and complicated, recursively generated management schemes. For instance, Amazon’s Mechanical Turks service puts together swarms of actual people to do repetitive, single-purpose tasks. The “Turkers”, a community of more than a 100,000 4
operate as “Artificial Artificial Intelligence” performing “Human Intelligence Tasks”.
Another cloud revolution in architecture is currently underway in the form of BIM. Building Information Modeling, among many changes it brings to the industry, is built on managing project changes as trackable transactions from groups of users that aggregate to a central file. This is the same “architecture” as the complex revision systems that computer programmers have been using for years to manage their source code. These “revision control” systems work by understanding the history of changes and dependencies embedded in any project. Large open-source projects have used this infrastructure to allow many programmers to collaborate on projects, creating their own “forks” that then get merged into the official code. Decentralizing the process of writing code is necessary in large open-source projects such as Mozilla 5
with contributions coming from weekend coders, teen-aged hackers, and professionals alike. Cloud infrastructures such as GitHub and Google Code have emerged to manage these transactions.6
Such services will eventually cater to architecture, with smaller outfits like SHoP Construction offering their own model-hosting services along with other BIM integration services7
to fill the vacuum. Whether efficiencies can be captured ultimately relies upon implementation, but the potential of these new modes of documentation will soon be necessarily tested when, as in almost all other media, the amount of documentation increases. What is less obvious is whether or not architects will be able to unlock the new work cultures of free agency and crowd-sourcing that these management systems allow.
Will the cloud reproduce architecture itself? Most of us already habitually frequent an alternate reality in the cloud - Google Street View. Street View seeks to build out a complete, explorable world through a gargantuan accumulation of geo-location and photographic data.8
In either case, both are encouraging users to spend more time in these artificial realities. Google even plans to keep the advertising updated on billboards so that no eyeballs are wasted. Part of this scheme includes the introduction of a complex system of bidding that will give buildings a similar value to the all-important “PageRank” that determines the visibility of web sites in searches.9
This is a characteristic business plan for future entrepreneurs in cyber real estate, and points to the both strange and familiar scenarios that will play out in the convergence of the physical world and the informational world. A clear cut dichotomy between the real and the virtual is becoming increasingly improbable as the world of information and the world of stuff increasingly resonate.This reciprocation created 800 petabytes of information last year, 10
most of it location-tagged and therefore spatially meaningful. By 2020, that amount will have increased more than 40-fold. 11
A increasing portion of this data will be generated from buildings and the array of sensors and monitors that are often found in new green construction. A large portion of this data might find itself on Pachube’s servers. Pachube (pronounced “Patch-Bay”) is already an invaluable tool to those interacting with the Internet of Things - IBM’s name for the network of low-power sensors and identifiers that are growing in numbers and utility.12
Countless sensors stream data to their servers 24/7, which aggregate it and serve it back in useful graphs and widgets. Through modern standards like ZigBee and Tron, these sensors form adhoc networks and are capable of cooperating to share information and create functionality. Cloud computing extends this level of cooperation into new dimensions, allowing building owners and occupants to analyze building performance over time and enabling researchers access to a massive quantity of information about the physical environment, updated in real-time.
"We can only access these potentials by accessing the cloud through its own terms and mechanisms - scalability and vast participation."
In our Cloud Lab introduction, we stated that clouds "bring new understanding to old problems" while simultaneously allowing us to pose new questions as well. However, we can only access these potentials by accessing the cloud through its own terms and mechanisms - scalability and vast participation. Tapping into these resources will require a continuous reformulation of the role of the architect in a world of data, users and informational processes - defining our own "human intelligence" tasks that create resonance between these diverse agencies. In the meantime, at least you wont have to wait so long for that rendering.
1 The architect even claims to have an honorary degree from Columbia University. http://wikipedia.org
2 From Reality Server Technical Overview
3 Interactive Visualization of GIS data with RealityServer. Urban planning in the city of Rotterdam. RealityServer is the ideal platform for centrally storing and visualizing architectural data throughout the entire design life cycle.”
4 NYTimes Artificial Intelligence, With Help From the Humans. March 27, 2007
5 Mozilla is the code project that powers the popular Firefox web browser.
6 “Gource in Bloom”, Demo of the new Bloom effect in Gource v0.21. Gource is a tool for visualizing user contributions to SCM systems.
7 SHoP Construction
8 Google Street View Behind the Scenes
9 US Patent “Claiming Real Estate in Panoramic or 3D Mapping Environments for Advertising“
10 2010 Digital Universe Study
11 “The Data Explosion” 281 Exabytes of Online Data in 2009 @ ReadWriteWeb
12 IBM Internet of Things