News

Appistry Rolls Out New Fabric for SOAs

• By Kurt Mackie
• November 9, 2007

EAF works across multiple computers in grid implementations, and the Affinity feature helps find data and run it on the associated machine, which improves overall system performance. It can partition data within an application fabric and cache dynamic data for quicker access.

The Affinity capability was developed by Appistry as it worked with GeoEye, a provider of satellite imagery that is used in clandestine programs and in MSN's Virtual Earth map service. GeoEye also provides satellite photos to the press.

"Affinity allows you to bring the data closer to the application and have that data live within a fabric," said Sam Charrington, Appistry's vice president of product management and marketing. "As a service request comes in to find the [GeoEye] image, Affinity can find that image on the exact machine, and the work can happen on that machine without moving the image around."

Affinity's Fabric serves customers with extreme transaction processing needs, as exemplified by GeoEye. The satellite-imaging company processes lots of data, and it is planning to launch its highest resolution earth-imaging satellite next spring. GeoEye's new satellite cameras will be capable of resolving a distance that's less than half a meter in length, Charrington said. Each of the images (about 20 gigabytes in size) will cover an area that's about the size of Texas, and the satellite will produce about five terabytes of data every single day.

While GeoEye may have the world's best photogrammetrists (image-processing experts), the company needed help on IT, Charrington said. Things like the OS-level programming and building out applications for multithreading, as well as distributed computing -- that stuff is hard for them, Charrington said, "and these guys are literally rocket scientists."

Appistry's EAF enables SOAs using grid computing, but the Fabric works differently than an enterprise service bus in an SOA.

"The way to think about a fabric and a bus is that a bus is communications infrastructure, but it doesn't really do a lot for the endpoints," Charrington explained. "And we are a platform for building the endpoints. So a bus will help you connect your services together, and we will help you build the services."

Appistry's technology features "scale-out virtualization," which lets developers and operators see the fabric connecting a grid of computers as one thing.

"The developer who is developing an application that is going to run on the fabric doesn't need to think about all of the traditional distributed computing issues -- which are the most complex issues in the development realm," Charrington said. "They simply develop their application as if it's going to run single-threaded on a single computer, and the fabric software virtualizes that application out across as many computers that are required for their needs -- it could be 10 or it could be hundreds."

Charrington described the Application Fabric as "the convergence of SOA, virtualization and grid."

Appistry supports customers that want to move away from the expensive grid computing model, running large servers for their SOAs. Charrington said that sort of approach can be costly when it's time to scale operations.

"When their business experiences a peak in demand and they need 20 percent more capacity, they've got to go buy another $100,000 computer to run their services," he explained. "Whereas the true notion of agility would mean just go buy some more blades or some inexpensive machines from Dell and just plug them in."

A more agile model is the one championed by Google, which is famous for building out its data centers using cheap x86-based computers, he said.