Current | 2005 | 2004 | 2003 | 2002 | 2001
2 June 2003
Profile of NGI-NZ Initiative
From Unlimited Magazine,
By Andrea Malcolm
They've obviously been reading Biotech Unlimited at Massey University. At least that's how I'm interpreting the latest moves at the Albany campus to create a grid-computing network. As I wrote in our last issue ("Grid formation", March), grid computing is fast becoming de rigueur among biotechnicians. By linking PCs or servers via the internet, technical gurus can create a powerful but cost-effective means of processing huge amounts of data.
Now Massey University is assessing the interest of other New Zealand organisations in establishing a grid. Chris Messon, who heads Massey's parallel computing research centre, says it would like to participate in such a way with other universities and research and development organisations.
Last year the university built a Beowulf supercomputer (see "Bioinformatics on a budget", Dec/Jan 2003). A Beowulf is a high-performance computer assembled from clustered PCs or nodes interconnected by a dedicated high-speed network and running an open-source operating system such as Linux or FreeBSD. Messon envisages Massey's Beowulf, dubbed Helix, would be linked via the grid to other Beowulfs and mainstream supercomputers. At this stage Massey is identifying other facilities with clustered supercomputers or those that are planning to build them, such as Hort Research and Auckland University.
Sounds exciting? Yes, except for one big problem: lack of a high-speed network around the country. What's needed is a thousand times faster than current average internet connection speeds. The US, Canada, Australia and the UK are already building or have built high-speed networks using technology known as Internet2 or next-generation internet (NGI). They're using NGI for interactive collaboration, real-time access to remote resources, shared virtual reality and large-scale, multi-site computation and data mining.
Scientists were instrumental in the formation of the internet. Nowadays, they need it for much more than email and discussion boards. They need access to huge data collections, very large-scale computing resources and high-performance visualisation for the individual user scientist. Called "e-science", these programmes rely on high-speed access to dedicated supercomputers or large numbers of low-cost networked computers, and to information stored in dedicated databases. NGI networks make it possible to deliver the required high computer processing capability by linking hundreds or even thousands of different processors in the grid-computing architecture. This is what Massey University is aiming to do.
In the UK, around £118 million has been invested during the past two years in the development of an e-science and grid-computing capability. Grids are also emerging to support large-scale geographically distributed meetings, seminars, lectures, tutorials and training. As higher bandwidth becomes more widely available, the compressed video systems that have evolved to combat a bandwidth-constrained world will progressively be replaced.
Aware that New Zealand could miss out on these developments, a number of organisations wanting to set up an NGI in New Zealand formed a consortium last October. Called the Next Generation Internet (NGI-NZ), it includes all the country's universities, Niwa, AgResearch, the National Library, Natural History New Zealand and InternetNZ (formerly the Internet Society). Each member has put $15,000 towards the project.
NGI-NZ was formed to carry out recommendations in a report called "Collaborating at speed" by the InternetNZ NGI steering group. This feasibility study, which examined the need for an NGI network in New Zealand, was paid for by a $50,000 grant from Industry New Zealand. More than 60 potential private and public sector users were canvassed, including the tertiary education and research community and industry sectors highlighted in the central government's growth and innovation strategy report - the creative sector, biotech and IT. The report recommends a network from Dunedin to Auckland with early extensions to other regions. User access would start at 1 gigabit per second with speeds up to 40gbit/s expected within two to three years.
Many other countries have been able to justify the establishment of a dedicated NGI for the research and education sector, largely because of their size. Use of these networks is governed by acceptable usage policies (AUPs) to exclude commercial and commodity traffic. New Zealand is probably too small for this approach to be realistic. Also the demand for bandwidth in the digital media and post-production sector already exceeds that required for research and education. Interconnecting with global NGIs will require some separation of commercial and non-commercial traffic at the international border, but this doesn't have to limit access within New Zealand.
Access to a global NGI will help ensure New Zealand-based researchers can link to their international colleagues. At a meeting in February, NGI-NZ approved a timetable aiming for a rollout of a national NGI network before the end of the year.
So what's holding back NGI-NZ? Our telco sector. It's not a matter of whether we have the technical capability but about compatible commercial agreements between the telecommunications companies. National providers are happy to support initiatives that will help bring about new network traffic and usage, but they're also concerned about the potential threat to existing revenues from applications that provide NGI users with a bypass opportunity.
Australia's high-speed research and education network, AARNet, is a fully licensed carrier in its own right and is able to offer its user organisations significant advantages by treating voice and data as one. AARNet has achieved its flexibility by seeking alternative suppliers of the main backbone technology infrastructure to avoid conflict of interest in reselling services from the traditional telecommunications providers. In Australia, Canada and the US this has been an option because of the healthy growth in competition and the supply of backbone wholesale bandwidth services.
New Zealand doesn't have this option. Wholesale services are available from metropolitan providers (CityLink and United Networks) and from international provider Southern Cross Cable Network (SCCN), but not between cities.
To connect to overseas NGIs, the Southern Cross Cable offers a range of capacity and pricing packages, and AARNet has also opened the door to a collaborative arrangement that would mean New Zealand could share its capacity. Access to global NGIs would then be provided via Pacific Wave in Seattle or the US Northwest GigaPop.
The technology used for an NGI in New Zealand will depend on the selected equipment vendors and broadband infrastructure providers. Given that both Telecom and TelstraClear have spare backbone fibre infrastructure, the report says a sensible solution may be to partner with one or both of these organisations to provide this component. This relationship would then determine other partnerships - routers, switches and so forth.
The difficulty with this approach is that both the major carriers compete vigorously in the retail market. This creates a serious conflict of interest in terms of negotiating attractive wholesale rates. Competitive issues would tend to dictate carrier behaviour, and while both companies are expected to offer big discounts for research and education, both will be wary of how this might creep to erode established markets.
This may prove insurmountable; the only choice would be to establish new infrastructure separate from the existing carriers. If the new entrant were to target the wholesale provision of long-haul capacity only, there would be no conflict with a private national research and education network. This model is now common in many other countries, including Australia, Canada and the US.
NGI-NZ is talking to telecommunications suppliers who have registered expressions of interest in providing network services. NGI-NZ chief executive Tone Borren says it is looking for a complete change in the way services are provided and charged for, recognising the change from a scarce resource to today's abundance of potential capacity, as well as trying to relate to the nature of LAN (local area network) users. It looks like interesting times ahead.
What's Internet2?
As you'd expect, Internet2, or I2, is a better version of the internet. Just like the internet, it arose from the work of US academics and government agencies. A consortium of universities started it because commercial use of the old internet was getting bogged down with too much traffic, thanks to Amazon.com book orders,
corporate emails, AOL chat sessions, spam and everything else. I2, which uses the latest in internet transmission software, namely IP Version 6, is strictly for non-commercial, serious research and development work.
Meanwhile, other countries have also started IP V6-based networks for their own scientific communities. Canada has CANARIE, Australia has AARNet, Singapore has SingAREN and the UK has UKERNA. These are all rapidly linking up through interconnection agreements. If you want to see which countries have one of these next-generation internets check out international.internet2.edu/partners/index.html. You'll note New Zealand isn't there.
|