While the US government funded TeraGrid project provides 10-30 G bit/sec interconnections between grid sites, nothing close to that is feasible in the commercial or enterprise grid world.  Not only would such high speed connections be prohibitively expensive for the enterprise customer, but not even a fraction of that bandwidth is generally available for nationwide connections - at any price!  

The enterprise grid user must try to match the 1/10 G bit/sec Ethernet connections in his data center and campus network with what is available in the metro and wide area networks.   Today, most commercial users interconnect grid sites using private lines- often by leasing DS3/E3 or dark fiber- at much lower rates than their campus LAN backbones.  But there are limits to connectivity and scalability with this approach and the burden of managing the network is on the enterprise. 

What role will the telcos play in grid networking? The driving factor for a telco grid network service offering will be to effectively use the assets it already owns and to realize a reasonable return on investment. What type of network service (point to point, IP packet, Ethernet virtual private line or virtual private LAN, etc), at what speeds, and with what QoS/ SLAs would generate the highest take rate, revenue and ROI for the telco? What price should be charged for such a service? Is there an elasticity of demand in such an offering?

Many other questions arise. When grids move from regional to national to global, how will they then be interconnected? Will it be a seamless evolution/ migration or disruptive?  When a single organization grid is opened up to collaborators, partner companies, and customers (as John Hurley of Boeing suggested) how will network access be achieved? Where will the enterprise grid networking specifications and best practices recommendations be developed? What type of network topology and connectivity arrangement works best for the majority of industry grids? Will that network technology adequately scale to accommodate more sites, more users, or more servers at any one site? What about disaster recovery? Finally, who will be the primary grid network provider - the user (or community of users), the grid hardware vendor (IBM, HP, SUN), or the network facility vendor (BT)?

Unless some answers emerge soon, grids are likely to be confined to a single campus network with 1/10Gig Ethernet links used for interconnection. This limits the size of the grid market, because it excludes interconnection of multiple, geographically dispersed grid sites. We are astonished that the telcos (with the exception of BT) are not more inquisitive and proactive in the emerging grid networking market and wonder if they are not missing out on a significant new source of revenue.