Distributed Network Processing
Written sometime around 1996 or 1997, by Jay Jacobson
As technology advances, and society becomes more reliant upon its advantages, clearly many new changes will have to take effect. However, it is rare for society to actually consider what these changes will be. People want technology to work for them, yet they are generally unwilling to make the sacrifices necessary to accommodate such advancements. After searching through all the hype, it is apparent that computer networking is likely to be the basis for our technological future. It provides us with the resources necessary to use cellular telephones, make a withdrawal from an ATM machine, create spectacular special effects for movies, receive satellite television and even communicate with people thousands of miles away. Unfortunately, due to massive popularity, computer networking has become slow, and at times, unsecured. Technologists have looked to several ideas for a solution, but most are nothing more than a temporary patch. While there are many possibilities to achieve a momentary resolution, most would be comparable to putting a Band-Aid on a severed limb. However, one new concept, called the MIPS Sucker theory, is a perfect fit in the lock of trouble. Although some would object, due to personal security issues, global implementation of the "MIPS Sucker" theory could ultimately revolutionize society.
The MIPS Sucker theory, commonly referred to as MST, is based on the current structure of the Internet. Started in the late 1960's, the Internet was originally called ARPANet. It was founded by the United States government to be a means of communication that would outlast a nuclear war. Its durability comes from the fact that there is no central point of operation; each computer connected to the network is fully self-sufficient. It was then "opened" to the public in the late 1970's. By the mid-1980's the World Wide Web (WWW) had started to form, and the Internet's popularity was becoming obvious.
The way it works is actually quite simple. There are two main components necessary to establish a computer network. The first is called a server which is simply a powerful computer with the ability to communicate with more than one external device at a time. The second component, called a client, is any computer with the ability to establish at least one connection with an external device. Utilizing the existing global network of telephone lines, multiple clients, each equipped with a modem, could contact and interact with a server (server "A"). As well, server "A" can also establish a link with another server (server "B"); giving all of server "A"'s clients a link to server "B" and all of its clients.
Unfortunately, a common networking problem is when a server becomes overloaded with its clients demands. This causes the entire network to slow down. As a result, all the users, no matter what they are doing, will see the effects. The servers of the Internet are constantly being upgraded in a desperate attempt to keep up with the mass amounts of electronic "traffic". This is where the MIPS Sucker theory becomes useful.
A MIPS is an acronym, which stands for Millions of Instructions Per Second. It is a measurement of how much information a computer can process in one second. When the clients' processes use all of a server's MIPS, any additional instructions for that server must be put on hold until the processing power becomes available. On the Internet, if a user tries to access a web page, that information is sent from, and processed by the server. However, many times the actual client's computer is sitting nearly idle, wasting it's MIPS, because the user is requesting information from a server. This "waste" is the fuel for MIPS sucking.
According to Laurence Vanhelsuwe, an independent software developer, the actual MIPS suckers are nothing more than small Java programs, called applets. If a server receives a large job to process, it could divide that job into small applets, then distribute those parts to its clients. The clients would automatically each process their portion of the task, and return the results to the server to be compiled into a finished product. Through the entire process, the individual user (client) would never notice a change. As well, a heavily loaded server could distribute portions to other servers that do not have as many clients (Vanhelsuwe).
For example, if a server had 10,000 clients, and it were given the task of calculating pi to 1,000,000,000 decimal places of precision, normally, all of the clients' requests would be put on hold, and their computers would sit idle, until the server reached an answer. Then, when it did reach an answer, it would have to process all the clients' requests that had backlogged. This causes a major bottleneck in the network. However, if the network were to use the MST, when the server received the request for pi, it could have divided that job into 10,000 applets (one for each client), then distributed them to be individually processed. Because the main job is now 10,000 times smaller, it can be easily processed by the clients' machines, and the server can be free to process the clients' other requests. When the clients' computers each finished with their portion of the answer, they would return it to the server where it would be compiled into the final answer. At this point, the server has kept up with the tasks assigned to it, the clients never noticed any performance decrease, and there is no bottleneck to deal with.
This form of raw processing power has been demonstrated many times by a non-profit group called "Distributed.Net." Their goal is to show the world that a million-dollar Cray with 2048 600MHz parallel processors and two gigabytes of RAM (per processor) is not necessary to crack the toughest computational tasks. Distributed.Net is conquering the same tasks faster than any computer ever built and they are doing it with 'average' users' computers spare cycles.
Although Distributed.Net does not use the same MIPS sucking software that is provided at the end of this report, it is based on the same principle. This group is currently working on an encryption project called DES II. The point of this project is to see how quickly anyone in the world can crack a 56-bit encrypted message. At the time of this writing, there are 18109 active nodes participating in the Distributed.Net group. Each of these nodes may represent an individual computer or an entire LAN. Through individual client-based MIPS sucking software, Distributed.Net is collectively processing approximately 459 billion 56-bit encryption combinations per second! Figuring that an average computer can perform approximately 58 MIPS, and half of that is idle time (while the owner is asleep, at work, eating dinner), that leaves 28 MIPS wasted.
Intel recently claimed that their super-computer could perform 1.8 trillion instructions per-second. Distributed.Net is cruising along, averaging 507.1 trillion instructions per-second. That is over 281 times faster than one of the world's fastest stand-alone super-computers. With that kind of processing power, it is no surprise that the participants of Distributed.Net have won every world-wide challenge that has been presented to them.
The MIPS Sucker theory would be favorable to society as a whole; from individuals, to corporations, even to entire governments. Electronic networking has become a vital part of today's society. With the Internet experiencing more than ten million users daily, it is obvious people have become dependent on networking. This dependence has caused many difficulties in the typical structure of electronic networking. This is because there are not enough resources available to support the demand. In turn, the Internet often becomes bogged down and unmanageable, as seen by the recent controversy between America Online and it's subscribers (HotWired News Network). However, the MIPS Sucker theory would have exactly the opposite effect. Rather than the number of users online causing difficulties, the network would run more smoothly with more users.
In an experiment conducted by researcher Bruce Krulwich, it was measured that MIPS sucker applets we able to "grab" about 40 percent of the processor power on a Windows NT client machine. This was still leaving the user enough processor power to launch other applications without noticing any decrease in performance (Schrage). "In other words, mipsucker-like applets are technically capable of transforming a Web site into a massively parallel meta-supercomputer that draws on the unused cycles (processor power) of visiting machines. Those previously unused cycles can be captured and fused together to perform all manner of computation." (Schrage).
MIPS sucking is so easily obtainable and versatile, it is like the solution that is so obvious, no one noticed it. With this kind of power at a programmer's fingertips, the possibilities are virtually endless. Using MIPS sucking applets, a moderately popular Web site, such as the HotWired News Network (http://www.hotwired.com) could actually give a programmer the power of thousands of SPARC and Pentium machines (HotWired News Network). A large corporation, such as Walt Disney, whose web sites receive over a million visitors per day, could use this technology to it's fullest extent. With one million visitors, and an average stay of twelve minutes, that adds up to 200,000 hours of processing time per day (Schrage). A corporation, such as Disney, could harness this power and utilize it to generate the special effects for their movies. Theoretically, this could save them millions of dollars that would otherwise have been invested in self-standing supercomputers (Vanhelsuwe).
According to Cray Research Incorporated, the world's leading manufacturer of super-computers, the parallel processing techniques used in the MST are very similar to those employed in the Cray T3D (Cray Research Inc.). These same techniques that have given the Cray T3D the title of "The World's Most Powerful Supercomputer" are obviously very effective (HotWired News Network). Yet, the MST has three distinct advantages over any stand-alone supercomputer. The first is the fact that the MIPS Sucker theory has no central point of operation, just like the Internet. This is advantageous because of the fact that when a portion of the network breaks down, the remainder of the network continues to function as normal. When the troubled part is repaired, it again becomes part of the global network without any interruption in service. Secondly, the cost for implementation of the MST is nothing. Because of the fact that the MST is based on the computers already on the Internet, and the MIPS sucker software is freeware (free, public domain software), the only requirements would be for the software to be put to use. Additionally, the MST has no physical limitations. Unlike the Cray T3D, which reaches it's maximum capacity with 2048 processors, the number of processors in a global MIPS sucker network is completely dynamic (Cray Research Inc.). Again, which means the more it is used, the more powerful it becomes.
This theory could actually create a form of "neural global network," allowing people and machines to interact as never before. This would speed up all forms of telecommunications, and virtually eliminate all "electronic bottlenecks" a computer networking environment. "Forget about the collective mind; the collective processor may be the future power." (Schrage). By people allowing their personal machines to support their own "electronic networking addiction," the MST could form an electronic society. This society, theoretically composed of billions of computers, could actually begin to form a neural network; supporting and maintaining itself independently. In consequence, people would have the ability to virtually travel anywhere in the world, and conduct any means of communication. All forms of telecommunications would be able to harvest substantial benefits from a global neural network. There would not be any more electronic "bottlenecks" because the number of pathways a communication could take to its destination would be theoretically endless. The MIPS Sucker theory could take us into the next major step in technology, similar to what the invention of the integrated circuit did for society.
The MIPS Sucker theory is extremely advantageous to society for
so many reasons. It is based on a platform that is already
established - the Internet. Because of the way electronic
networking had integrated itself into society, MIPS sucking would
provide a seamless solution to the problem of not enough
resources to support peoples' electronic demands. The power that
could be harnessed by a MIPS sucking network is almost
unimaginable. It could be used for such a wide range of
purposes; anything from conducting long distance business deals,
to attending school, to creating special effects for movies, even
computing pi. A close relative of the MST is already in use in
the world's most powerful supercomputer and the world's fastest
and most powerful computational device is an actual MIPS sucking
network. Why not let the world see the advantages MIPS sucking
could offer? It could give society an entirely new perspective
on life. One where the electronic problems of today are a thing
of the past and distributed networked processing is the wave of