Informatics Practices · Ch 2 — Emerging Trends
Grid Computing
Grid Computing
Suppose a research problem needs more computing power than any single affordable machine can supply. One answer is to make many ordinary machines act as one extraordinary machine. That is grid computing: a grid is a computer network of geographically dispersed and heterogeneous computational resources — machines of different kinds, sitting in different places — working together, as shown in Figure 2.13.
How a grid differs from a cloud in purpose
The cloud's primary focus is to provide services. A grid is different: it is application specific. It exists to attack a particular large problem, and in doing so it creates the sense of a virtual supercomputer with enormous processing power and storage — even though no single supercomputer exists anywhere in the network.
Nodes: the building blocks
The constituent resources of a grid are called nodes. These nodes temporarily come together to solve a single large task and reach a common goal; once the goal is met, the temporary association can dissolve.
Why is this practical today? Countless computational nodes — from hand-held mobile devices to personal computers and workstations — are already connected to a Local Area Network (LAN) or to the Internet. Most of them sit idle much of the time, so it is economically feasible to reuse their spare resources, both memory and processing power. The grid therefore lets scientists solve computationally intense scientific and research problems without procuring costly hardware — the capacity is borrowed from machines that already exist.
Two types of grid
- Data grid — used to manage large and distributed data, providing the required multi-user access to it. The emphasis is on where the data lives and who can reach it.
- CPU or Processor grid — the emphasis is on computation: processing is moved from one PC to another as needed, or a large task is divided into subtasks that are allotted to various nodes for parallel processing.
Grid computing versus the IaaS cloud service
The two can look similar — both give you computing capacity you do not own — but the arrangement is fundamentally different:
| IaaS cloud service | Grid computing | |
|---|---|---|
| Who supplies the capacity | A service provider who rents the required infrastructure to users | Multiple computing nodes that join together |
| Relationship | Commercial: rent and use | Cooperative: pool resources for a common goal |
| Purpose | Whatever the paying user wants to run | A common computational problem shared by the nodes |
Middleware: the glue that builds a grid …
The diagram sits inside a rectangular frame and has a single box at its centre labelled "Grid Resource Management System". This central box is the coordinator of the grid — the component that decides how the pooled resources are put to work.
Around it, arranged in two rows, are the machines that make up the grid. The top row shows a database server, a desktop workstation, a tablet and a laptop; the bottom row shows a printer, a smartphone, laptops and a rack server. Every one of these devices is joined to the central management system by a double-headed arrow, and both rows carry the same caption: "Users sharing their resources". …