Feature - Wireless grids: Squeezing a grid onto a widget
As wireless devices become increasingly common, and common devices become increasingly “smart,” wireless grids become increasingly practical. That means that the timing is perfect for WiGiT, a wireless grid testbed which will begin testing its alpha software in June.
The purpose of WiGiT (Wireless Grids innovation Testbed), according to the Syracuse University project leader Lee McKnight, is to refine open specifications for a wireless grid standard, and create a stable platform for experimentation.
“With WiGiT we expect to be able to do these large scale experiments from campus to campus, and we can run little experiments on that,” McKnight said. “Open specifications will make it easier for others to latch on.”
WiGiT, a National Science Foundation-Partners for Innovation program-funded collaboration between Syracuse University, Virginia Tech, MIT, and Tufts, will not have to start from scratch. Various research projects have already laid the foundations upon which WiGiT will build. In 2002, the NSF funded a project called “Virtual markets and wireless grids.”
They started by looking at dynamic edge grids – a grid across a mobile phone, a laptop, and a few peripherals, for example. The goal was to determine what would be involved in creating that sort of grid, and what market mechanisms would be involved. The result was that, in 2004, patents were filed, and a spin-off company was formed – the Wireless Grids Corporation.
Meanwhile, there was a middleware development project in progress.
“We were taking open source mobile devices, and we were trying to jam the Globus Toolkit onto that and it just wasn’t working,” McKnight said. “That was instrumental in helping us understand that you can’t really get there from here.”
Instead, they decided to focus on creating an edge grid for now, connecting back to the big grid world at a later date.
“We know that smart devices will become capable, but we know we don’t need the whole toolkit,” McKnight explained. Instead, they focused on creating a stripped down, custom toolkit for wireless edge grids. Today, beneath the wireless grid, there’s a software-defined radio infrastructure called CORNET, developed by the Cognitive Radios and Networks group at Virginia Tech. The CORNET project emphasizes cognitive engine design, self-organizing networking algorithms, and network security.
Wireless grids have a wide variety of potential applications. Environmental science, for example, is likely to be among the first areas of application, because of the potential uses for large networks of wireless sensors. To achieve that potential, researchers will have to overcome a number of limitations, working within a large set of constraints.
Power is a major concern, because mobile devices generally operate on batteries or in some cases, solar power. The limited bandwidth available via a wireless connection is also a concern, as is storage space for any data which is recorded. Finally, mobile devices have limited processing power.
“We’re talking about sort of smart and dumb devices at the edge,” McKnight said. The smart devices are faster, with more storage space than the dumb devices. Even so, they are limited in what they can handle in comparison to the nodes found on a traditional grid.
The smart devices can dynamically manage the dumb devices, coordinating the transfer of information, data, and computational jobs. It could transfer power-hungry jobs from devices running low on power to fully-charged devices. It could transfer more computationally intensive jobs from dumb devices to smart devices, and from there to a traditional grid node.
“There’s this ad hoc layer of smart devices that could make these decisions,” McKnight said. “Sometimes it makes sense to do it locally. Sometimes it should be passed along. And that depends on what the purpose is. And it also depends on the dependability of the intermediary device resource near the sensor.” Security and encryption are also areas of difficulty; some security measures may overload the device, making it difficult to transfer while keeping the information safe.
McKnight is looking forward to studying the data they record once researchers begin to use WiGiT for their own experiments.
“We’re going to have this incredibly huge data set of our early applications and users. Analyzing that will I think be very interesting and we’ll learn a lot,” McKnight said. “So our purpose is to first do experiments on real resources, real networks, and with real people, and second… agree on open specifications, new standards, so that anyone who wants to play can say okay, here’s the open APIs that are recommended by WiGiT.”
The WiGiT team will be able to use the data they record for more than just troubleshooting their technology. It will also be invaluable for learning how people interact with wireless grids.
It’s still early days for the WiGiT project, which began only a year ago with a funding grant from the NSF-PFI. This summer, the Syracuse WiGiT team plans to coordinate with some of the experiments running on CORNET at Virginia Tech. They’ll also be working on layering new applications over their lightweight grid edgeware. By fall, they aim to release the first draft of the WiGiT specifications.
Researchers from a number of institutions have recently joined the collaboration, including the University of Virginia, University of Wisconsin, Rochester Institute of Technology, and the State University of New York College of Environmental Science and Forestry.
“We expect to have the process for others to join in more formally set up by July,” McKnight said. “If you join us, you’re consenting to share data on what you do with WiGiT with us. We’re studying you or your use of WiGiT.”
What next? The WiGiT collaboration is expected to produce open specifications by 2012, but McKnight is expecting the fruits of their labor to live on and evolve. Said McKnight, “We have a bandwagon. We want everyone to jump on our specification.”
—Miriam Boon, iSGTW