Green Supercomputing

Sorry for the double-posting, but I don’t want Professor Gus to have to read up on our blog over Winter Break, and I really feel this subject should be put forth to the community.

When an industry consumes large amounts of any resource, it soon finds itself in the crosshairs of environmentalist groups, as it well should. Cyberscience is no exception to this, as data centers absorb staggering amounts of power. In class we have already covered the progressive response of many companies to move to “greener” energy sources. Yes, that is a very useful step to take, but until solar, wind, and water power becomes efficient enough to power all of the world’s data centers, this cannot be the only step taken. Luckily, many companies share my view on that.

There is a list of green supercomputing sites called the Green500 List, on it I was surprised to see many government-owned computers. I knew the government has need for many data-centers, but our nation isn’t the most environmentally friendly country out there. Apparently, some scientists got to work real hard on making these Computing centers as eco-friendly as possible. So I decided to probe further, and found out I didn’t have to look much further.

If you were to google “green supercomputing” right now, you would first see a link for the Green500 list. The very next link is concise, yet very well written article about NASA’s use of green supercomputing. The center they are speaking of is called “Pleiades”. The first thing they mention is its power efficiency. They speak proudly of their #54 spot on the Green500, but even more proudly of their #5 spot in efficiency(232 megaflops/watt). The entire system was built from the ground up with efficiency in mind, and their initial thoughtfulness has proven very beneficial in the long run. They also have special programming to idle unused machines within the system, and perform constant maintenance on the entire system to keep it operating at full efficiency.

This system is a great example of green supercomputing and its ability to be equally as powerful as regular supercomputing. Its great amount of efficiency has saved NASA much money (don’t worry, they will find somewhere else to spend it), and more importantly, has kept its carbon footprint extremely low(for a data-center).

Google “green supercomputing” if you want to view the articles referenced in this post, they are results 1 and 2.

To the Cloud! Cyberscience’s mentality makes its way into our daily lives

In the formative years of computer technology, computers were something you came to when you had a task to get done. Computation was something to be done in a centralized location. With the rapid evolution of computing technology, personal computers became the norm for most computing. There always remained a need for large-scale computing centers, and the rapid evolution of these centers has made such things as “The Cloud” possible. Now computation could be done off-site, with results displayed on-site, for everything from playing a video game to doing truly high-throughput computing.

The Cloud is a very ambiguous term, generally referring to anything involving off-site data storage or computation. The simplest implementation of this would be programs along the lines of Google Docs. Google Docs is a program that stores and runs documents. Documents aren’t very computationally or data intensive, but the mentality of not having to store things on ones own computer is the main point of their use. Many internet sites offer off-site data storage services as well. Then there are programs like VNC. VNC (Virtual Network Computing) allows one to set up ones own personal computer to be remotely accessed by another device over the internet. Cloud computing has even advanced to the point where a company has found it better than distributing physical gaming consoles. OnLive allows people to play the latest videogames over broadband internet. No, this isn’t an ad for OnLive, personally, I like the option of playing games without requiring broadband hookup. However, one cannot deny the merits of a system that can (according to some reviews I read) play advanced games seamlessly while being controlled from many distant locations.

In summation, the technology for “The Cloud” is quite obviously improving at a rapid rate. The situation one might foresee is one of returning full circle, in a way, back to data being computed at centralized places. With the ever-lowering cost of technology and the explosive growth of “Cloud” technology, coupled with the ever-expanding area of broadband internet availability, it seems entirely possible that all personal computers could be bare bones machines with only enough functionality to connect themselves to an off-site location. If it were cost-effective, who wouldn’t want a light as air machine that would run for extended periods of time, and that has limitless computing potential?

Cyberscience and the Humanities: Not Just Dan Katz

I know that at this point Google Books is well-established, but this article I read in the New York Times reminded me of our class and gave me some new interesting thoughts about the fourth paradigm and the humanities:

In 500 Billion Words, a New Window on Culture

The article offers a glimpse at how the ability to search billions of words from an astounding number of documents can influence research about human culture.  Using Google’s database,  researchers of “culturomics” can look at how culture has changed, ranging from languages to the length that fame lasts.    It presents the idea that large databases have the ability to influence understanding of humanistic topics much more intensely and quickly than previous means, and that it is beginning to catch on throughout academia.

When Dan Katz came to class, I think that we were all intrigued by how influential cyberscience can be on the humanities.  I think we all saw it as kind of a rare occurrence.  However, this article offers another window into how the fourth paradigm is changing even the humanities very rapidly.

The most impressive notion that Dan Katz described was that there are researchers at the Law School still leafing through books one page at a time, one pair of eyes at a time, and that it seems a very silly way to go about doing research involving so many words and ideas.  This article seems to me to discuss a very similar issue, in which research that recently took months of time for very  few results now has a high tech avenue that can pay huge dividends.

I don’t doubt that we will be seeing many similar articles in the future, as the fourth paradigm seems to keep popping up in the most interesting (and sometimes surprising) places.

Computational science at home

Recently I came across a program called BOINC.  BOINC is a program that you can install on a home computer and it collects data from a central server.  It then takes the data and uses your home computer to do computations with that data.  When it is done it sends the results back to the central server.  BOINC is a way of using peoples home computers to sort through data instead of requiring a supercomputer to provide quick calculations on massive amounts of data.  Currently I have BOINC running on a Linux home server for a cosmology@home project.

Clearly BOINC has its disadvantages, but for something like this to be available to underfunded scientists shows massive potential.  This is a big part of computational science in regards to infrastructure.  Instead of spending a lot of money or waiting a long time to gain access to a supercomputer, scientists can use a network of home computers whose processing is given freely.  The upside of BOINC is that often a computer is sitting around doing nothing, so why not dedicate that unused processing power for a positive benifit.  Programs like this help to expand the field of cyberscience, and create new paths for the field to move into.

 

http://boinc.berkeley.edu/

http://www.cosmologyathome.org/

Flocks, Swarms, and Computing

After watching all of our class’s presentations last week as well as agonizing over my own (in a good way 😀 ), I must say I found the ‘Theoretical Ecological Modeling’ group’s topic rather interesting as it made me think how the behavior of living things can be modeled and simulated by computers. Although we normally don’t notice it, the social structures that other organisms on earth form are amazingly complex and remain understudied. From the flocks of birds that carpet bomb the Diag as they make their annual migrations to ant colonies that are somehow able to build underground cities, many animals somehow ‘know’ how to perform feats that we humans would consider extraordinary for their lack of intelligence and technology. Finding some way to model and analyze these complex behaviors through computing is a fascinating new area in science.

How such individually simple organisms such as insects come together to form an intelligent and goal oriented society or ‘hive mind’ still remains a mystery. Though we know that many animals communicate via powerful chemical signals or pheromones, the complexity involved in organizing hundreds to millions of individuals is mind-boggling. It would be cool to see a computer model show how a bee colony organizes workers to seek out plants to pollinate, care for its young, and defend against intruders. I’ve also wondered how animals such as bird flocks or schools of fish migrate over distances of thousands of miles through treacherous environments in one piece without any explicit sort of planning or strategy that a ‘smart’ human would do. Perhaps as computers become more powerful, we could simulate variables such as local climate and food supply in a realistic model of animal migration.

The prospect of computers modeling the adaptive behavioral patterns of animals in nature also reminds me of Michael Crichton’s sci-fi novel Prey. Admittedly, it was a cheesy, and scientifically absurd story of how a laboratory accidentally releases a swarm of intelligent nanorobots into the desert. The nanorobots were programmed with a ‘predator-prey’ algorithm that gave them a collective intelligence that gave the swarm the ability to learn, reproduce, and evolve eventually giving rise to predatory behavior against humans. The thought of computers become more like living organisms is both an interesting and frightening possibility as computing power grows exponentially. The ability to make large numbers of individually insignificant and simple ‘bots’ work together to perform large scale projects from medical treatments to space exploration may one day become reality. However, as artificial intelligence becomes more and more in tune with the vicious and relentless nature of natural selection and predator-prey relationships, this technology may become dangerous.

Either way it makes for somewhat entertaining science fiction.

 

Reaction to Computational Legal Studies

After Dan Katz’ presentation about Computational Legal Studies, the magnitude of the possibilities hit me. So far in legal studies, paper and brainpower has been the driving technology. In general, law requires much more thought than computation so it makes sense that computers would not have too much of a place. The only problem is, humans inherently have two problems: they are never perfect, and they have a limit as to how fast they can process information. For example, A computer can read a book in a fraction of a second while it takes a human several hours at least. Computers also almost never make mistakes. Of course they can fail, but it is almost always the fault of the human who wrote the program that failed. Computers aren’t the solution yet; they sill have one major shortfall, the ability to think. AI has always been out of reach, but with major advances in technology, it is not out of sight. There can never be a true AI, but when computers have the ability to make basic logic decisions, they could be a very powerful tool in law. Imagine a machine that can read in a speech made by an opposition lawyer, analyze what the argument point is, cross reference against a database of laws and precedents, and come up with specific counterarguments in a matter of seconds. These then could be used by a lawyer to refute the opposition’s point. Nowadays, it is up to the lawyer to memorize all of the precedents and laws and prepare for the case beforehand, but with technology like this, useless mass memorization would be avoided and lawyers would be able to focus on the actual logic part of the argument. My guess is as good as anyone’s as to what the future holds, but the possibilities are almost endless. One thing iis certain though, computers will, one day, influence the law process.

Racetrack Memory

A team at Texas A&M University has found a way to improve the speed and efficiency of racetrack memory by “using a series of current pulses rather than DC, AC, or a combination of the two.” While this new model of memory storage is still far from replacing hard disk drives or solid state drives, this is a step closer to making the new technology practical.

Racetrack memory works by storing memory on a thin wire as magnetic domains that can be pushed across the wire with pulses of current. If it becomes a practical alternative to hard disk drives, it will improve supercomputing with its faster read and write speeds. Since it also has a higher data density, it will make the current deluge of data take up less space in our data centers. Because it involves no moving parts, it is more energy efficient and produces less heat, making data storage more green.

More about how racetrack memory works

Constructing an Effective Cyberinfrastructure

Dan Atkins presentation focused on the evolution of cyberinfrastructure from the early days of computing until now. His presentation provided us with a precise definition of the concept as “a platform of technology and human and organizational support for advanced, integrated computation and information resources in the service of research and learning.” Breaking this definition down, I see cyberinfrastructure as the system and the ease with which it allows access to digital data to anyone wishing to.

Of course, with the vast amount of data available and the ever-increasing rate at which it is being created and collected, the challenge is to make access to it comprehensive and practical enough to be convenient to use. That is where Dan Atkins defines his concept of a cyberinfrastructure ecosystem, where all the components work cohesively together. This includes expertise, networks, and organizations such as universities, libraries, and research centers to coordinate with each other and the scientific instruments, data, software, and computational resources available. He emphasized the necessity of scientists to become data-savvy as data Is becoming an increasingly critical part to scientific research (hence the fourth paradigm). For this, truly investing in an ecosystem approach to cyberinfrastructure will prove necessary.

The use of environmental terms in discussing cyberinfrastructure isn’t merely a quirk. The ideas of sustainability and stewardship mentioned by Dan Atkins most effectively capture the nature of what we are facing when it comes to managing data. Just as we must manage our natural resources and protect the environment for future generations, data needs to be sustainably maintained in its explosive quantities for future researchers and scientists to utilize them efficiently. In fact, Dan Atkins even provided a concrete example on its relevance from the UK, an article stating that the growth of the knowledge based economy relies on a national cyberinfrastructure that will support the scientific community with easy transfer of knowledge. The idea of a nationwide and standardized cyberinfrastructure, in fact, seems to be an efficient method of organizing this vast machine into an easy-to-use unit.

I found the three-ring model from his lecture to be helpful in envisioning the issue better since it showed the interconnectedness of actual research and learning, operating the cyberinfrastructure, and providing supportive systems and environments for future e-scientists. With one link weak, the other two follow.

The School of Information that Dan Atkins helped found in the early 1990’s is an example of a possible solution to both the logistics and politics on cyberinfrastructure, especially since it consists of legal experts, economists, and computer scientists. This embodies the collaboration he repeatedly emphasized as necessary.

Structured Query Languages and their use in cyberscience

At the end of Chris Miller’s presentation he asked if any of us were familiar with databases at which point no one in the class raised their hands.  He then told us of the importance of SQL databases and their high level of use in cyberscience.  I realized that the other guest lecturers had mentioned the use of databases, and began to wonder what exactly they were in the literal sense.  So after class, my interest piqued, I looked up more on what SQL was and why it was so valuable.  After a quick search on google I came across several pages on SQL which I found stands for System Query Language.  Essentially SQL is a computer language used to store and search for tables of data, and SQL servers are servers that host and distribute this data.

There are many different types of SQL services available such as Microsoft SQL Server and MySQL, but they all tend to use the same framework and command interface.  They are readily interfaced with other programming languages such as PHP and Python which have built in methods to query them and request data based on specific parameters.  This is an important part of SQL because this means that the entirety of the data can be stored on a SQl server and all the programmer has to do is connect to this server and request the data as the program computes whatever task it is designed for.  In effect many different people can work with the same data at the same time without the need to have all the data on their machine.  Instead they can simply request pieces of data as the program calls for it.

SQL also makes searching for particular data much simpler.  A program can send a command to a SQL server like SELECT * FROM table_name WHERE (condition).  The program can specify a condition for the data to be sent back. For instance, if there is a table of students and the program only needs student whose first name is Steve the program can send SELECT * FROM Students WHERE firstname = ‘Steve’ and the database will send the information regarding all students with the first name Steve.  This saves time on both ends and means particular programs don’t have to download all the data and sort through it all only for the data it needs.

Databases with common commands like this are obviously important perhaps mostly in regards to consistency and simplicity of design.  SQL databases do seem to go against the idea of bringing computations to the data, but they make it much easier to share data which is very important to the growth of cyberscience.

Computational Legal Studies Regarding The Citation Network

It is no great secret that lawyers cite other decisions while in court, anyone who has seen a courtroom drama can attest to that. Actual lawyers would emphasize the power of citation even more. In its basest form, law is simply a decision made based upon fairness and morality, as agreed upon by the general populace. This is a simple concept, but to make every decision based upon ones own feelings would land a judge in hot water in no time. Laws are made, and laws are agreed upon, and it is common practice to navigate a moral grey area with citations of similar cases. So in order for the law to function effectively, numerous citations need to be collected, organized, and made available to the law community. This is one of the things that Dan Katz is working on. During his lecture he showed us a three dimensional rendering of early court decisions and their citations. They began with few citations, but as time went on citation grew exponentially. Court decisions started to form clusters of related decisions and the most cited cases were made to move toward the center of the screen. All in all, it was a very interesting presentation.

After enjoying the aesthetically pleasing video of twenty years of court decisions, I mentally took a step back and assessed the show from a technological point of view. This had been a clear-cut example of cyberscience. The sheer amounts of pertinent data each case must contain already puts it into high data crunching levels, but the cases are all searchable, using some sort of algorithm to determine which cases are pertinent to the searched criteria. One would assume such a thorough algorithm would take a tremendous amount of computing power, and barring the instance that a prodigy of programming wrote it, it most definitely is a case of computationally intensive science. In a field almost completely built off of citation this is a much needed and infinitely useful tool.

For anyone interested in watching/re-watching the video, I have included it here.