/* */?> /* */?>

Português (Brasil)English (United Kingdom)

Latest News

  • Patch 2011-09-29
  • 09.22.2011 - Version 4.2.6 released
  • 08.16.2011 - Version 4.2.5 released
  • 07.15.2011 - Version 4.2.4 released
  • 03.11.2011 - Version 4.2.3 released
The OurGrid Vision PDF Print

The recent advances in computing and networking are changing the way we do scientific research, a trend that has been dubbed e-Science. Thanks to the power of computer-based communication, research is now a much more collaborative endeavor. Moreover, computers play an ever-increasing role in the process of scientific discovery. Data analysis without computers sounds antediluvian. Simulation has joined theory and experimentation as the third scientific methodology. At the same time, processing large amount of data is becoming increasingly popular in the industry as well. As a result, many IT users now demand non-trivial computing capabilities .

As a solution to this demand, Grid Computing has appeared with the enticing promise of turning computing into utility. As the technology matured, several different types of grid infrastructures have been built and are currently in operation. These, in turn, may be broadly divided in two classes, namely: service grids and opportunistic grids.

Service grids normally assemble high performance, dedicated computing resources, such as clusters, supercomputers, and large data storage systems that are spread over a relatively small number of administrative domains. For instance, one of the largest service grid currently in operation is the EGEE grid which encompasses tens of thousands of processing cores distributed among approximately a few hundreds of sites. Service grids provide high and well defined levels of quality of service (QoS) that is achieved thanks to a lot of effort being placed in the monitoring and management of the infrastructure. On the other hand, opportunistic grids are somewhat more “lightweight” grid infrastructures that scavenge idle computing cycles from non-dedicated resources, providing a best-effort high throughput computing infrastructure.

OurGrid is an open, free-to-join, cooperative grid in which sites donate their idle computational resources in exchange for accessing other sites' idle resources when needed. It uses a peer-to-peer technology that makes it in each site's best interest to collaborate with the system by donating its idle resources. OurGrid leverages from the fact that people do not use their computers all the time. Even when actively using computers as research tools, researchers alternate between job execution (when they demand computational power) and result analysis (when their computational resources go mostly idle).

For OurGrid to be useful, it must be fast, simple, scalable, and secure. These were the four major goals that guided the design of OurGrid. Putting it in more detail:

 

  • OurGrid must be fast (i.e. the turnaround time of a job must be much better than what is possible using only local resources) otherwise there will be no point in using it.
  • Simplicity is also a fundamental requirement for OurGrid. After all, users want to spend the minimum possible effort on the computer technology that will solve their problems. They want to focus on whatever research they do. Computers are just tools for them.
  • OurGrid must scale well, otherwise it will not tap the huge amount of computational power that goes idle in the sites around the world. Note that scalability is not just a technical issue. It also has administrative ramifications. In particular, it is not acceptable to have to go through a human negotiation to define who can access what, when and how (something that is needed to set up service grids). Therefore, OurGrid must be a free-to-join open grid.

  • OurGrid must be secure, because its peer-to-peer automatic granting of access will allow unknown foreign code to access one's machine. Nevertheless one's machine must remain safe.

 

Achieving these goals was a very challenging task. In order to simplify somewhat the problem, at least for now, we reduced OurGrid's scope to supporting Bag-of-Tasks (BoT) applications. BoT applications are those parallel applications whose tasks are independent. Despite their simplicity, BoT applications are used in a variety of scenarios, including data mining, massive searches (such as key breaking), parameter sweeps, simulations, fractal calculations, computational biology, and computer imaging. Assuming applications to be BoT simplifies our requirements in a few important ways. In particular, we can deliver fast execution of applications without demanding any QoS guarantees from the infrastructure. It also makes it easier to provide a secure environment, since network access is not necessary during the execution of a foreign task.

In short, by focusing on BoT applications, we were able to deliver a useful solution for the compute-hungry users around the world. In fact, OurGrid is in production since December 2004. Any site can join the system by downloading and installing the software. Note that no human contacts or negotiation are needed for a new site to join the system.