Grid computing is a phrase in distributed computing which can have several meanings:
- Multiple independent computing clusters which act like a "grid" because they are composed of resource nodes not located within a single administrative domain. (formal)
- Offering online computation or storage as a metered commercial service, known as utility computing, computing on demand, or cloud computing.
- The creation of a "virtual supercomputer" by using spare computing resources within an organization.
- The creation of a "virtual supercomputer" by using a network of geographically dispersed computers. Volunteer computing, which generally focuses on scientific, mathematical, and academic problems, is the most common application of this technology.
These varying definitions cover the spectrum of "distributed computing", and sometimes the two terms are used as synonyms. This article focuses on distributed computing technologies which are not in the traditional dedicated clusters; otherwise, see computer cluster.
Functionally, one can also speak of several types of grids:
- Computational grids (including CPU scavenging grids) which focuses primarily on computationally-intensive operations.
- Data grids or the controlled sharing and management of large amounts of distributed data.
- Equipment grids which have a primary piece of equipment e.g. a telescope, and where the surrounding Grid is used to control the equipment remotely and to analyze the data produced.
During 2007 the term cloud computing came into popularity. It is conceptually identical to the canonical Foster definition of grid computing below. In practice all clouds are grids, but not all grids manage a cloud.
Grids versus conventional supercomputers
"Distributed" or "grid" computing in general is a special type of parallel computing which relies on complete computers (with onboard CPU, storage, power supply, network interface, etc.) connected to a network (private, public or the Internet) by a conventional network interface, such as Ethernet. This is in contrast to the traditional notion of a supercomputer, which has many processors connected by a local high-speed computer bus.
The primary advantage of distributed computing is that each node can be purchased as commodity hardware, which when combined can produce similar computing resources to a multiprocessor supercomputer, but at lower cost. This is due to the economies of scale of producing commodity hardware, compared to the lower efficiency of designing and constructing a small number of custom supercomputers. The primary performance disadvantage is that the various processors and local storage areas do not have high-speed connections. This arrangement is thus well-suited to applications in which multiple parallel computations can take place independently, without the need to communicate intermediate results between processors.
The high-end scalability of geographically dispersed grids is generally favorable, due to the low need for connectivity between nodes relative to the capacity of the public Internet. Conventional supercomputers also create physical challenges in supplying sufficient electricity and cooling capacity in a single location. Both supercomputers and grids can be used to run multiple parallel computations at the same time, which might be different simulations for the same project, or computations for completely different applications. The infrastructure and programming considerations needed to do this on each type of platform are different, however.
There are also some differences in programming and deployment. It can be costly and difficult to write programs so that they can be run in the environment of a supercomputer, which may have a custom operating system, or require the program to address concurrency issues. If a problem can be adequately parallelized, a "thin" layer of "grid" infrastructure can allow conventional, standalone programs to run on multiple machines (but each given a different part of the same problem). This makes it possible to write and debug programs on a single conventional machine, and eliminates complications due to multiple instances of the same program running in the same shared memory and storage space at the same time.
Design considerations and variations
One feature of distributed grids is that they can be formed from computing resources belonging to multiple individuals or organizations (known as multiple administrative domains). This can facilitate commercial transactions, as in utility computing, or make it easier to assemble volunteer computing networks.
One disadvantage of this feature is that the computers which are actually performing the calculations might not be entirely trustworthy. The designers of the system must thus introduce measures to prevent malfunctions or malicious participants from producing false, misleading, or erroneous results, and from using the system as an attack vector. This often involves assigning work randomly to different nodes (presumably with different owners) and checking that at least two different nodes report the same answer for a given work unit. Discrepancies would identify malfunctioning and malicious nodes.
Due to the lack of central control over the hardware, there is no way to guarantee that nodes will not drop out of the network at random times. Some nodes (like laptops or dialup Internet customers) may also be available for computation but not network communications for unpredictable periods. These variations can be accommodated by assigning large work units (thus reducing the need for continuous network connectivity) and reassigning work units when a given node fails to report its results as expected.
The impacts of trust and availability on performance and development difficulty can influence the choice of whether to deploy onto a dedicated computer cluster, to idle machines internal to the developing organization, or to an open external network of volunteers or contractors.
In many cases, the participating nodes must trust the central system not to abuse the access that is being granted, by interfering with the operation of other programs, mangling stored information, transmitting private data, or creating new security holes. Other systems employ measures to reduce the amount of trust "client" nodes must place in the central system such as placing applications in virtual machines.
Public systems or those crossing administrative domains (including different departments in the same organization) often result in the need to run on heterogeneous systems, using different operating systems and hardware architectures. With many languages, there is a tradeoff between investment in software development and the number of platforms that can be supported (and thus the size of the resulting network). Cross-platform languages can reduce the need to make this tradeoff, though potentially at the expense of high performance on any given node (due to run-time interpretation or lack of optimization for the particular platform).
Various middleware projects have created generic infrastructure, to allow diverse scientific and commercial projects to harness a particular associated grid, or for the purpose of setting up new grids. BOINC is a common one for academic projects seeking public volunteers; more are listed at the end of the article
CPU-scavenging, cycle-scavenging, cycle stealing, or shared computing creates a "grid" from the unused resources in a network of participants (whether worldwide or internal to an organization). Typically this technique uses desktop computer instruction cycles that would otherwise be wasted at night, during lunch, or even in the scattered seconds throughout the day when the computer is waiting for user input or slow devices.
Volunteer computing projects use the CPU scavenging model almost exclusively.
In practice, participating computers also donate some supporting amount of disk storage space, RAM, and network bandwidth, in addition to raw CPU power. Since nodes are apt to go "offline" from time to time, as their owners use their resources for their primary purpose, this model must be designed to handle such contingencies.