Collider piques interest in improving Energy Sciences Network
The activation in September of the European Organization for Nuclear Research's Large Hadron Collider might produce an embarrassment of riches.
The riches would be in the wealth of data that will be generated by the world's largest particle supercollider. The embarrassment would be the struggle to transmit via digital networks to laboratories around the world, such as those operated by the Energy Department.
To get this data to scientists, DOE will be using its Energy Sciences Network, which has a capacity of as much as 100 gigabits/sec via a backbone provided by Internet2. But making the best use of that raw bandwidth can be a challenge, said Brian Tierney, leader of DOE's Advanced Technologies Group at Lawrence Berkeley National Laboratory.
'We think we're ready for it,' Tierney said shortly before the LHC was fired up for the first time. 'We've been getting ready for it for some time.'
The paths for LHC data have been determined and the network links and endpoints using them have been fine-tuned to handle the flows. But DOE also wants to optimize ESnet performance for other applications used by scientists with little expertise in networking. ESnet has been experimenting for about six months with Phoebus, a technology that adds new protocols that negotiate optimal network services for an application.
'It seems to be pretty solid,' Tierney said. 'The performance improvement can be quite dramatic,' especially over long distances. It shows greater improvements on a coast-to-coast transfer than from state-to-state. 'But there are a couple of things missing before we could deploy it in a serious way.'
The first issue is making it easy to use. After all, these are physicists, not network administrators. 'It would need a Windows interface,' instead of the current Linux setup, Tierney said.
Security is another issue. 'That's a hard problem, to figure out what the right model is,' he said. Scientists probably would not want to be bothered with X.509 certificates, so the Secure Shell protocol might be the answer. 'We haven't given a lot of thought to this yet.'
Tierney has been working with Phoebus developer Martin Swany about needed changes in the system. He is also an assistant professor of computer and information sciences at the University of Delaware. But in theory, Phoebus should improve current conditions, in which some scientists use USB drives and sneakernets for large file transfers because the current 100 gigabits/sec network works too slowly for them.
'On today's high-speed networks, it should only take eight hours to move a terabyte,' Tierney said. 'It takes longer than that to load a USB device.'