In Fukuoka, Japan, a specialized operating system, using graph-theoretic methods and primed for the next generation of exascale systems, is being developed to study extremely complex interdependencies in urban settings.
The InfiniCortex project aims to create a global InfiniBand fabric to connect supercomputers across continents for these tasks and already has nodes in Singapore, Australia, Japan, USA and Poland. The network is expected to include several new partners in Europe and circumnavigate the globe by the end of 2015.
Dr Marek T. Michalewicz, the Chief Executive Officer of A*STAR Computational Resource Centre in Singapore will be overseeing a public presentation on the topic, called “Understanding Urban Development through HPC” on July 13 at the International Supercomputing Conference in Frankfurt, but we thought it would be useful to glean more information about the project in advance.
Sieslack: Could you explain the concept of the “Smart City” and what it offers?
Michaelwicz: The adjective “smart” has become a platitude. It has nearly lost its proper meaning as nowadays almost anything that functions automatically is labelled as smart. To me “smart” has some flavor of: capable of prediction, pro-active, autonomous, self-correcting, such as capable of reacting and adjusting to stimuli never experienced before …
In that sense it’s hard for me personally to call things smart, even though they may be enabled for complex functions or function on an auto-pilot mode. However, when the system is complex and dynamically responds to external stimuli and adapts to new scenarios, the situation changes. This might very well be the case with cities.
But, there must be a mechanism allowing for a feedback. – This is furnished by a pervasive system of data sensors and transmission networks, combined with models which are fed with this data, and of course vast computational resources for processing and storage.
I can refer to a paper “Instrumenting the World” (Estrin et. al., 2001) and another one by Intel (2004) under almost the same title, which described possibility, importance, pervasiveness and global expansion of sensor networks. We should only look into incredible strides and developments in sensors, sensor networks, micro electro-mechanical systems (MEMS), and all related areas, to see that the vision of “Smart Cities” might actually be close to real. I have, myself, painted such a vision in my talk “Brave New World of (nano)-sensing: the next technological revolution” at a Nanotechnology conference in Cairo, Egypt in 2009.
The Smart City concept in a nut-shell, is a city very densely instrumented with sensors and sensor networks measuring the state of this very complex system, transmitting the data for real-time or near real-time processing, and having mechanisms in place to react, according to some pre-determined measures. Typically you want to measure things like the power grid and electrical energy utilization, ,as a function of time, location, concurrent events, fuel prices, etc.; metropolitan mass transport systems, as a function of time, location, the calendar and special events, correlation with road traffic, etc.; human traffic and pedestrian thoroughfares; noise and pollution; local weather patterns, heat island effects, and related mitigation systems; water supplies and demand, public safety and crime prevention, and so on.
Sieslack: What kind of remote sensor and computing infrastructure will be required to realize this?
Michaelwicz: There are multitudes of sensor technologies already available to conduct most of these continuous or discrete measurements. There are technologies to monitor usage and economy of energy consumption within the buildings, at electrical power grids and along the roads. There are several categories of sensors and sensor networks among the most important being wired and wireless.
For example, many sensor networks within the buildings or along the major infrastructures, such as bridges, roads and highways, can be wired. This simplifies the matter since wires serve the dual purpose of conducting signals and providing electrical power to the sensors.
Wireless sensors can be more widely dispersed, and can even be embedded in structures such as bridges and road surfaces for monitoring their health. The difficulty lies in powering them. There are new methods of power harvesting, super-capacitors, piezoelectric micro-power generators, etc., but the problem is still open.
Another almost trivial example of a very dense and pervasive sensor network is the network of mobile devices carried by most of us. We might not know it but each new mobile phone or portable “smart device” has the ability to measure a number of different attributes including temperature, the magnetic field, pressure, acceleration, geographic location (GPS), etc. We are very far from utilizing all data available from this massive scale network.
Sieslack: What kind of software tools will be needed?
Michaelwicz: This is one of the great challenges in this new field. One needs to see the whole vertical stack: controllers in the sensors and embedded systems, transmission protocols, data storage, data analysis tools, and new paradigms for correlating model predictions.
In a sense it parallels some of the older military or scientific sensing scenarios, such as a navy sonobuoys systems, seismic networks, GPS systems, or weather observations stations But for Smart Cities we see an extremely high density of collection points, perhaps bandwidth saturation, and very high sensor mobility, There is also a wide variability of environmental conditions to contend with. – For example, phenomena such as human stampedes, flash floods, terrorist attacks, and car crashes on major thoroughfares are critical elements that must be monitored.
Some of the models will be characterized by extreme computational complexity and will rely on sophisticated methods of statistics, data mining, Bayesian inferencing, graph theoretical methods, combinatorics and others.
An example of specialized software might be CityOS which will be described during our ISC session by professors Fujisawa and Matsuo. Another example is SALI artificial intelligence, a rule-induction technology being developed by Alan Boyd in China.
Sieslack: Is there work taking place today that would serve as a foundation for Smart Cities?
Michaelwicz: There are many intense programs focused on creating Smart Cities, I can’t even list them all. Let me start with one that will be featured at our special session at ISC15.
In Fukuoka, Japan, a group from Kyushu University is developing a special Urban OS to manage the entire chain of data, including collection at the sensors, transmission, cloud resource provisioning and storage. This work will be reported by professors Katsuki Fujisawa and Hisato Matsuo.
In Chicago, the Urban Center for Computation and Data (UrbanCCD) a joint initiative of the University of Chicago and Argonne National Laboratory under the direction of Charlie Catlett, is developing the “Array of Things”.
Yet another, unprecedented, intense and massive scale effort is seen in Singapore, where, on the 24th November 2014, the Prime Minister Lee Hsien Loong has launched an official National program of “Smart Nation”. The government agency responsible for implementing this vision is the Infocomm Development Authority of Singapore, which embraced this task with vigor, under a slogan: “Many Smart Cities – One Smart Nation”. To date many separate research entities are engaged in parts of this program: including the Complexity Institute at the Nanyang Technological University the Future Cities Lab, a joint enterprise of National University of Singapore and ETH Zurich. Also contributing to the effort are government research organizations such as the National Environmental Agency (EDB) (weather, climate, noise pollution, heat island effect, haze), Housing Development Board (heat, human traffic, energy issues), Land Transport Authority (human movement, road systems, Mass Rapid Transfer system) and my own A*STAR, where a Complexity Research Program is run within the Institute of High Performance Computing (IHPC). The research done by the IHPC Complexity Program will be presented at our ISC15 by Dr. Christopher Monterola, who directs this effort.
Through my private contacts I have unofficial information that China is planning to initiate a spectacular program of building 300 new “greenfield” Smart City projects that encapsulate cities, districts of existing cities, towns, and even villages that will eventually house over 300 million people. They are currently screening available technologies and working on standardization to create a uniform system of cities. The scope and size of this program will overshadow all other programs taken together around the world. The Ministry of Housing and Urban Development (MOHURD) is charged with the task of screening the technologies.
One important piece of information is that we will see a widespread use of InfiniBand, beyond the datacenter confines, as THE fabric to carry information and data from all sensors and common home appliances to the two types of datacenters, one dedicated to computing and the other for storage.
Sieslack: What is the next step toward supercomputing and smarter cities?
Michaelwicz: These developments have the life of their own, but the HPC community needs to be observant to the yet-to-be-required massive scale of computational demands needed to support Smart Cities. Depending on the success of our session, we might think of featuring this topic as a standard item on the ISC conferences program. We may also expand the scope of our work to the more general topic of sensors and especially sensor networks, not necessarily in the context of Smart Cities, but as another area with serious computational demands.
We hope to see you all at ISC 15 from July 12-17 in Frankfurt, more about this talk can be found: http://www.isc-events.com/ct.php?id=174