Road to grid computing remains difficult
Subscribe now for $100 (23 issues) and save more than 37% off the cover price!
Get the latest news from Computerworld delivered via email.
Sign up now
Grid and cloud computing’s real heyday is yet to come, says grid pioneer Ian Foster.
Reliable infrastructure management for remote applications and data is still difficult enough to put a large “potential energy barrier” between the providers of utility computing services and potential users.
“It takes some substantial effort to still make use of these services or to become a service provider.”
However, there are encouraging signs; already scientists and businesspeople are calling on services provided by several remote computer facilities and combining them.
“If you look at what people are doing with computing in this part of the real world, they’re not simply invoking programs that may run locally or remotely — they’re often performing very complex activities that may involve data from one location, software from another,” Foster said in an Australasian Computer Science Week keynote.
New Zealand-born Foster is now based at the US Department of Energy’s Argonne National Laboratory and at the nearby Chicago University.
“Providing services that can support those workloads in an on-demand manner is a reasonably complex activity,” he says. It could involve more than one grid.
There is, for example, a Social Informatics Grid, which lets researchers browse a large resource of social sciences data. Having chosen the data and formatted it appropriately, the researcher can shunt it off for analysis to the Open Science Grid, which carries the right tools and a large enough resource of computer power.
The aim of utility computing, Foster says, should be to provide compute-power and access to data and software as electricity is provided; a supply elastic enough to meet unforeseen peaks in demand.
The Cancer Biomedical Informatics Grid (CaBig) enables the cancer research community to share data more effectively. Mutual awareness of research and collaboration at a distance has historically been done by peer-review publication and by a letter or a disk in the mail, says Foster. Instead, on CaBig, researchers are encouraged to publish data and software as services using a service-oriented architecture.
This grid supports the sharing of a wide variety of data types using security and authorisation mechanisms. They are based on those developed by the Globus consortium (a joint effort of Argonne and the University of Southern California) “but applying in this case to information and software” rather than the computing power and storage that Globus’s toolkit was originally designed to administer, says Foster.
Over the past two years the business world has begun to make an effort to reduce the energy barrier, he says, by providing very simple services that meet the demands of a large number of people in the industry. This is “the cloud”, a metaphor that encompasses a variety of different ways of using remote computing.
Software as a service is the best known use of the cloud, but reliable, available infrastructure itself can also be provided in a service mode, or platforms. Amazon has developed such a service for its own use, known as Dynamo, and is now providing it to outside users.
Unfortunately, says Foster, these platforms are currently proprietary in their design and interoperability is currently being tackled.