Grid Computing: Future Development

Wong Lok Kwan

Grid is an infrastructure that enables heterogeneous geographically separated clusters of processors to be connected in a virtual environment (network). Grid computing is a technical means to solve problem by sharing and applying the resources of many computers in the network. There are three main reasons that make grid computing popular. The first one is making more cost-effective use of a given amount of computer resources. Besides, it is a good way to solve problems that can't be approached if not enough amount of computer power is provided. And the last is that it suggests that all the resources of the member computers can be cooperatively harnessed and managed as collaboration toward a common goal.

Due to its many advantages, some models are developing and designing in progress for grid computing. Examples are as follows. 

Corba, it is a set of facilities linked through “off-the-shelf” packages. It is a client server model using web based technology. And it wraps the existing code used in Java objects. What is more, it utilizes current public key security techniques. Furthermore, it also exploits thread level and object level parallelism.

Legion, it is a single unified virtual machine that tying a large number of objects and processors together. This is object oriented machine. Each object defines the rules for access so that the object interfaces will be well defined. In addition, there is a core set of objects which provide basic services. So, the users can create and define their own objects easily and in a comfortable way. Besides, legion works with high performance, users can first consider the load and job affinity then select the hosts. The presence of object wrapping characteristic help to support the parallel programming. And because the user autonomy exists, users can choose security arrangements as well as the scheduling polices.

Globus, it is an integrated ”bag” of basic grid services. Its middleware layered architecture structure helps to builds global services on top of core local services. Apart from that, translucent interfaces to core Globus services, well defined interfaces that can be directly accessed by applications. In addition, its system can be improved like incremental implementation or other existing tools can be enhanced or replaced if needed.

In fact, grid computing has been promoted as the global computing infrastructure of the future for several years. Some scientists even considered grid computing as one of the major sources of the impact to technological and scientific changes on the society and economy, because of the fact that the usage of large amount of data becomes crucial to many aspects.
However, there are still some challenges existing, including the latency limitations, improved protocol schemes, and additional grid-based tools are needed.

But the most important thing is the cost problem. In order to amortize the substantial costs of managing these large data volumes (the key property of computing grid), the cost will usually be shared by scientists. Besides, large IT system is required to exploit the huge amount of data volumes. However some disciplines with no experience in administrating and managing these complex systems will be affected. These users can neither afford to establish a sufficiently large local compute nor to manage suitable IT systems by their own. Although the significant investments in the grid concepts, it recently draw less attention. Instead, some new concepts like cloud computing seem to replace it.

In fact, computer users will give up if the expected benefits dwarf the transition costs. In this way, we have to enhance the both sides. Many of the publications enhance the eventual benefits, but at the same time should we focus on the transition costs for the user? Can we identify user groups who want to use grid computing so that they may help us to make the system becomes mature then to reduce transition costs? However, this approach will focus on the core services and may create barriers in building a grid with higher level services. But we should consider carefully that do most users actually need such a high level services grid?

In short, there are many advantages of grid computing which is worthwhile to use it. But the inevitable fact is that there are still many problems should be solved first. If those problems are solved, grid computing will be one of the best techniques.