What is grid computing?
The general principle of grid computing consists in the availability of a network that connects geographically spread computing and storage resources while giving many user groups access to this network. Each user can gain access to the totality of the resources (computing capacity, memory, software, storage, etc.) that have been added to the network by other members of the network. Grid computing means in fact a globalisation and virtualisation of computer infrastructures.
The term “grid” comes from the concept of “electricity grid” that is available in the industrialised world and where all power plants are connected to each other in a grid structure. A power shortage in a power plant is immediately and transparently replaced by power of another power plant in the grid.
Another image of grid computing is the idea that the required computing- and storage resources can be got immediately by connecting to a network.
In theory all types of computers and storage equipment can be added to a grid infrastructure.
Does grid suit your research activities ?
Grid is best suited for applications with a data-parallel nature where the work can be split up easily in multiple independent parts (embarrassing parallel jobs).
Typical grid applications are :
- Massive data processing workloads
- Large computational job collections that require a minimal time to completion
Due to its capacity to accommodate a variety of research groups from different scientific disciplines and to connect resources across national, organizational, administrative… boundaries, the Grid infrastructure is well suited for projects that requires collaboration and resource sharing with national and international partners.
It is almost evident that the connection of computers and storage to a network, e.g. internet, is not sufficient to realise a computing/data grid. There needs to be software that has the task to make sure that all systems that are part of the grid infrastructure can be announced to the grid users at any moment and with all their properties. There need to be a “distribution system” that attributes free resources to users that request those resources. Rules that control the access to the grid or part of the grid or that can give priority to certain user groups have to be defined. The necessary security measures have to be taken when giving users access to the grid. That type of software is called grid middleware.
Cloud computing is very close to grid computing in the sense that both share the same fundamental idea of having access to resources through a simple Internet connection. Cloud computing could be seen as an evolution of grid computing that integrates all the new possibilities offered by advanced technologies such as virtualization for example. The major differences with grid stands in a few characteristics : wide variety of types of resources and services offered, flexibility, on-demand resource provisioning, fast and ubiquitous access to resources, automation and orchestration,…
Further reading : Cloud Computing and Grid Computing 360-Degree Compared