Electrical computers and digital processing systems: multicomput – Distributed data processing
Reexamination Certificate
1999-01-07
2002-07-09
Coulter, Kenneth R. (Department: 2154)
Electrical computers and digital processing systems: multicomput
Distributed data processing
C709S235000, C709S239000, C709S238000, C709S203000, C712S028000, C712S029000, C705S014270, C705S037000
Reexamination Certificate
active
06418462
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates in general to distributed computing, and more particularly to parallel computing and web computing. This method is disclosed first in a paper entitled “Global Sideband Distributed Computing Method” on Jan. 13, 1988 at the Western MultiConference (WMC'98) by the Society for Computer Simulation (SCS).
2. Description of the Related Art
1) In the Internet, there are a lot of CPU resources that are far from being fully utilized
The computer industry is a fast growing industry. Every day faster computers are introduced to the world. From the Intel 8086 to the Pentium, CPU speed has more than tripled. However, the utilization of CPU power still can not keep up with the growth of the computer chip industry. As a result, there are more and more idle CPU cycles in computers in the world. For example, a typical office computer would be idle 80% of the time and most home computers are used only as Email and news reader or word processors that consume less than 30% of the computing power for a modem Pentium CPU.
On the other hand, the need for larger computing power is not decreasing. For example, complex differential equations, neural network simulations and other design tasks at big entities such as GM or NASA certainly have strong demands for huge computing power. One way to fulfill those demands is a supercomputer which is 10 to 100 times faster than a desktop PC. However, the expense is so high that it is normally beyond the budget. Another good solution is distributed computing. However, traditional distributed computing has some limits that are not easy to overcome.
Since there is much more idle CPU power worldwide, if we can develop a method to utilize it, we can save a lot of wasted processing power. This invention discloses a new method—a global sideband service distributed computing method to achieve that goal.
2) Traditional distributed computing has a lot of limits
Normally, in distributed computing, we can access more computer power, memory and I/O bandwidth by connecting several machines together. Groups of machines can provide high performance more cheaply than traditional supercomputers. Based on this concept, people have developed several distributed systems such as PVM, MPP or recently CORBA and JOE.
However, there are some limits that are uneasy to overcome. For example, one limit is the maximum number of computers in a group. PVM sets a maximum number of computers at 4095. Some methods are based on a local network. As a result, it is difficult to enable distributed computing in a worldwide environment.
Another prominent dissatisfying limit is accessing each participating computer. For example, in PVM, if a user wants 1000 computers to work together, he has to physically locate them and obtain accounts from each node. It also limits the scalability in that no idle computer in the world is allowed to join the distributed computing group freely.
Pat. WO 98/00780 entitled “Client—Server Technique for Distributed Computing” disclosed a method for remotely running server applications through a web interface. It is useful for client administration and metering of application usage but is not a general distributed computing method. U.S. Pat. No. 5,740,362 entitled “Management of Network Distributed Agents in a Distributed Computing Environment” disclosed a method providing a single access point for resources distributed within a network but traditional limits still apply.
In summary, the premise of a traditional solution is that a user already has direct access to all the participating computers. This requires the user either physically owning the computers or having to pay for their use. Therefore the expense is not ignorable and sometimes is not practical. Furthermore, it doesn't make use of many idle computers scattered around the world.
The method disclosed here, to some extent, can overcome these limitations. It is significantly different from traditional distributed computing methods.
3) network servers can attract enormous numbers of clients
As the Internet is growing hotter, more computers are getting connected. In 1993, there were 2,152,000 hosts on the Internet and it is estimated to have 100,000,000 in 1998. Together with the invention and development of many new tools and technologies such as WWW, CGI, JAVA, CORBA and so on, Internet services are also becoming common to users. In principle, any host that offers some service can be regarded as a server, like a web server, gopher server, news server, chat server or game server.
Most network servers are designed on a client—server basis in which clients connect to a server, begin a conversation to get information or perform an operation. This conversation is also called a transaction or session. Usually, the server is the information provider and the client is the receiver. The goal of the service provider is to attract as many users (clients) as possible.
Because of the infinite size of the Internet, popular servers have many clients. For example, very hot web servers such as the Yahoo search engine could be accessed nearly 10,000 to 1,000,000 times daily. It is also common that at a given time, a game server could have thousands of clients implementing a fighting or mud game for hours per day. Statistics show that during the games between Deep Blue and Kasparov last year, the news web page of IBM's Deep Blue was accessed more than 1 million times per day.
Here we can see that good servers entice many users to access, thus there is an opportunity for the server to distribute tasks to them during conversations. This is the basis of our sideband service method.
4) Sideband service is another channel of conversation between client and server
Under normal situations, the server provides information, service or something else and the client receives them passively. If the server asks the clients to perform some tasks for it, then the role for both sides is reversed and usually another separate conversation must be set up. This type of connection only exists when clients connect to a server for the service it requires and cannot exists by itself. The term sideband channel (or separate dedicated communication channel) is derived from this scenario.
Since most computers on the Internet have quite huge amounts of idle CPU time, it is natural that the server can use it for parallel computing. The only limit is that the tasks the client can perform shouldn't influence the overall performance of the client computer, nor should it take too much time (longer than normal conversation). The client might need to transfer back results to the server. Therefore by assigning different tasks to different clients, the server can acquire huge amounts of computer time since the number of clients can be very large (more than 1 million per day in some cases).
In sideband service, the relationship between client and server is reversed in the process of computing. For sideband service, clients performing some task for the server may be regarded as a dynamic server since it serves as a “computing server”.
5) Benefit of the sideband service computing method
There are at least two benefits of sideband service: one is that if many many clients work together, we can achieve something that is hard or impossible by a single computer. The other is that commercial value can arise from sideband service, as it could be regarded as a service charge to maintain the normal network service.
In summary, in sideband distributed computing, the server can ask all clients to work for it at the same time the server provides the clients some service. Clients gain access to the service and in return, they pay back with part of their idle computing resource. It is the computing power contributed by all clients that makes up our whole distributed system.
6) No other methods to date can make use of idle CPU resources on the Internet while being acceptable to the public.
Some other methods such as network advertisers exchange their services with advertising. Clients ha
Coulter Kenneth R.
Xu Yongyong
LandOfFree
Global sideband service distributed computing method does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Global sideband service distributed computing method, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Global sideband service distributed computing method will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2854381