OurGrid is an open-source grid middleware based on a peer-to-peer architecture. OurGrid was mainly developed at the Federal University of Campina Grande (Brazil), which has run an OurGrid instance named "OurGrid" since December 2004.^[1] Anyone can freely join it to gain access to large amount of computational power and run parallel applications. This computational power is provided by the idle resources of all participants, and is shared in a way that makes those who contribute more get more when they need. Currently, the platform can be used to run any application whose tasks (i.e. parts that run on a single machine) do not communicate among themselves during execution, like most simulations, data mining and searching.^[2]

Overview

The main structure of OurGrid

The OurGrid software is written in Java which provides the cross-platform support, i.e. any operating system which can run the Java virtual machine can participate in the grid. It consists of four parts: Broker, Worker, Peer and Discovery Service. The Broker is used when the user needs to use the grid for some computation. The Worker is used when the user doesn't need to compute anything at the moment but wants to provide idle computation resources in order to gain the reputation in the network. The Peer is used when the user controls multiple machines and it allows to control the connected Workers. The Discovery Service allows multiple Peers to interact and exchange their computational resources.^[3]

Workers in the OurGrid support virtualization which provides security for the users. For example, some malicious users could upload a task which connects to the Internet and thus organize a DDoS attack, but when the task is ran in the sandboxed environment, it is isolated from the host's file system and can't connect to the Internet.^[3]

Incentive mechanism

To prevent some users from getting large computational resources "for free" (i.e. only using the grid for computation and not providing any computational resources in return), OurGrid uses the mechanism called the Network of Favors. When the user provides idle computation resources for someone else to use, they gain reputation in the network, and the more reputation they gain, the more prioritized their queries will be when they request some computation from the grid in the future.^[3][4]

The Network of Favors assumes that all users are eager consumers, i.e. every user seeks to obtain more computational resources. The user's reputation is not stored somewhere globally (e.g. in the central server), but is stored locally by the peers who directly interacted with the user. The reputation never becomes negative, otherwise malicious users could just create a new identity with a clean reputation.^[4]

Limitations

In 2013, Marek Šimon, Ladislav Huraj and Vladimír Siládi analyzed the performance bottlenecks of such P2P grid applications as OurGrid. They concluded that not every parallel task can be effectively solved in the network (more specifically, the tasks for which the overhead of distributing data between the workers is comparatively large). For instance, they devised a task which used interpolation methods for determination of the snow cover depth. They measured the time it took for the sequential algorithm to solve the task and compared it to the time it took for the grid to solve parallelized version of the task, and there was no increase of efficiency due to the large overhead for data distribution and collection. However, the other application which deals with radioactive decay showed the increase of efficiency due to the large volume of data, more complex computations and small overhead of data distribution compared to the computation itself.^[5]

References

External links

• Official website