Ian Foster's view of cloud computing and the grid

In the recent news of "Cloud Computing", Business Week published an article titled "Google and the Wisdom of clouds", followed by Ashlee Vance comments on the article, Nick Carr also wrote a book about cloud computing, "big Switch." So what is cloud computing? What's so special about it compared to other technologies? What new changes can it bring to it?

The basic idea of cloud computing is that in the future computing does not have to be done on the local computer, or on a centralized device that is manipulated and controlled by Third-party computing and storage operators, which reduces the number of times to unzip and install the software.

Of course it's not a whole new idea. In fact, as early as 1960, the founder of Computing, John McCarthy, predicted that one day in the future computing would be organized as a public facility, and that the way it was implemented was predicted.

By the mid 1990s, the term grid was used to describe technologies that allowed consumers to acquire computational power on demand. I (Ian Foster) and some other peers have pointed out that by developing a standard protocol for requesting computational power, we can construct a computational grid that is similar to a grid in form and application. The researchers then expanded and extended the idea in various ways, developing a number of large-scale integrated systems (federated system) such as TeraGrid, Open science grid, caBIG, Egee and Earth system grid. These systems can not only provide computational power on demand, but also provide data, software and other information. Standardization organizations such as OGF and Oasis have also established some relevant standards. In large numbers, the industry grid and cluster are equivalent, but there is no real commercial grid computing provider, at least not recently seen.

So can "Cloud Computing" be simply considered a new name for grid? In the IT field, it may take up to five years for technology to escalate or update itself, so it is difficult to give a definitive answer to the question (but we can think about it from different perspectives).

It can be considered that the two are equivalent because they have the same perspective or starting point, and they want to make the use of the computer by their own purchase and operation to be provided by the third party, in order to reduce the computational costs, enhance the reliability and flexibility of computing.

But it can also be said that there are differences between the present situation and a decade ago compared to a big difference, the current application of massive data analysis and processing to a large extent, increased the demand for computational power. While recognizing the benefits of switching from mega machines to commercial clusters, we also find that managing and operating these clusters requires expensive costs. On the other hand, we have low cost virtualization implementation technology. And most importantly, companies such as Amazon, Google and Microsoft have spent hundreds of millions of of billions of dollars on building a commercial grid that contains tens of thousands of machines. And as long as a credit card can be achieved. On-demand access to hundreds of thousands of machines across hundreds of data centers around the world is tempting (these resources are appropriate to address some of the problems that need to be addressed on a large scale and potentially distributed data)! So it can be said that the size of the data manipulated by the two is different, and that the operation of these new large-scale data may require some completely different methods to deal with the related problems. At the same time, it may also solve some new problems that have not been discovered, and many times it is true.

But the problems that grid and cloud need to solve are largely similar. There are some common requirements: the need for large equipment management, the need to define methods to support users to discover, request and use the resources and services provided by the centralized device, and perform some highly parallel computing tasks on the appropriate resources. Although they differ on a number of details, the two communities are working to solve some common problems.

Unfortunately, some of the methods and techniques used to achieve these goals in the commercial cloud computing system have so far not been open and generic. On the contrary, are basically developed and managed by some companies to facilitate their own use. The idea that we want to implement interoperability between different providers in the grid (like power grids) has not been reflected in cloud computing, and the fact that grid technology and related protocols clearly explain the problem should not be overlooked.

Many people end up with the view that we see the same market in cloud and grid. The emergence of the first cloud computing cluster (cloud computing clusters) is very similar to the grid cluster of a few years ago. Perhaps the advent of Oracle 11g will bring a new opportunity for the combination and development of the two.

In the future, which one is more promising? Cloud or grid? I'm here to make some bold predictions, and I think the computing economy will become more and more like the energy economy. However, tomorrow's energy and computing economy will not be the same as yesterday's power grid. With small-scale producers coexisting with large regional producers (wind, solar, biomass, etc. in the energy economy), there is an increasing number of local clusters and embedded processors in the computational economy, and the continuous improvement of dynamic load distribution between them, I believe that the energy and computing economy will be in the direction of the combination of micro-products and large equipment. I know that although the development of computational power is unlikely to go exactly the same way as the development of electricity, I firmly believe that we can still see the similar evolution and development process driven by similar demand.

In the process of building distributed grid or Cloud (Groud), integrated virtual systems on demand must be able to provide the necessary computing power and performance according to end-user requirements. We need to define protocols that enable users and service providers to discover and distribute requirements to other providers, monitor and manage their resource reservation mechanisms, and complete the appropriate payment management. We also need relevant tools to manage the underlying resources to implement distributed computing. We need both a centrally managed cloud computing device and a distributed and interoperable grid device.

Some of the future necessary protocols and tools will come from Amazon and Google staff, some from researchers dedicated to the grid, and some from grid and cloud, whatever we call (perhaps Groud) researchers. In a sense, it is a fascinating thing for these different communities to develop along parallel lines or eventually find a common destination.