Feature - Joining the dots: creating interoperable grids

Interoperability is now making grids bigger and more able than ever before. Images courtesy of OSG (top) and Marijke Unger (bottom)

Pooling computers into grids is enabling scientific discoveries that would not otherwise be possible.

Always looking to the future, grid developers are now asking: is there a compelling case for connecting these grids together?

As grid usage grows and more researchers discover grid power, many predict that scientific appetite for this power will also grow.

Already satisfying large appetites as individual ventures, two U.S.-based grids—Open Science Grid and TeraGrid—agree that interoperability between their infrastructures is the next step.

The interoperative ideal

To a scientist, grid interoperability should mean nothing beyond reliable and continuous access to highly responsive computing and data storage resources. These resources, provided by one or any number of computing grids, could ideally respond interchangeably to user requests.  

To the providers of grid resources and services, interoperability means coordination, adaptation and compromise—especially when the two interoperating grids are as different as OSG and TeraGrid. In this case, interoperability begins with establishing basic levels of commonality and agreements.

Thinking to the future, these two grids have already begun moving towards the interoperative ideal.

The nanowire simulator on nanoHUB.org can send jobs to both Open Science Grid and TeraGrid, automatically selecting a grid depending on the parameters of the run. Last year, some 650 users launched more than 6,700 simulation jobs using this particular tool, which is one of 56 tools available on nanoHUB.org. Over the last twelve months nanoHUB served 25,000 users worldwide. Image courtesy of nanoHUB

Introducing the grid hybrid

“One key element of the first phase of OSG–TeraGrid interoperability has been our agreement to align on the same version of Condor and Globus,” says OSG middleware coordinator Alain Roy, referring to the base software for both OSG’s Virtual Data Toolkit middleware and TeraGrid’s CTSS middleware. A second key element, says Roy, was TeraGrid’s adoption of OSG’s build and test software.

“As OSG adds patches, TeraGrid rebuilds Condor and Globus to stay in sync,” adds John-Paul Navarro, Roy’s counterpart at TeraGrid.

TeraGrid’s software is split into an essential core kit with optional feature kits, each of which contains some subset of the Globus services, so this achievement represents more than meets the eye.

“We’re enabling users to draw on both sets of resources using the same basic remote interfaces,” says Dane Skow, Director of TeraGrid’s Grid Infrastructure Group. “Already a few communities are building infrastructures which exploit this. We’ll have to see where the demand leads, and determine what level of integration is needed to make this more broadly attractive.”

Under the hood, Open Science Grid and Teragrid have the same engine, but a very different look and feel. Used variously as a Ferrari, Mac truck, commuter vehicle and special-occasion Cadillac, the grid hybrid is already satisfying new customers.

- Anne Heavey, Open Science Grid