I am having trouble making sense of the article “Coming Soon: Superfast Internet” (http://www.timesonline.co.uk/tol/news/uk/science/article3689881.ece) that appeared in the Times Online. The article doesn’t make clear what it is about the Internet that is being altered. That is the perplexing part. Cloud communications is powerful. New applications will emerge. Much more bandwidth will be needed. Latency will be an issue. I get all of this–but this is nothing new or profound. It is the continuation of the Internet roller coaster we have been on for the past decade. Why would the article emphasize an obsolete Internet? Journalistic sensationalism is part of the answer as it makes for great headlines. Let’s look past this though.
I have not done research to gather more information–and my knowledge of TCP/IP and other underlying Internet technologies is limited. Maybe readers can help shed light–in the meantime, I will offer some speculation.
Let start with the definition of the Internet. Using wikipedia: “The Internet is a worldwide, publicly accessible series of interconnected computer networks that transmit data by packet switching using the standard Internet Protocol (IP). It is a “network of networks” that consists of millions of smaller domestic, academic, business, and government networks…”
Relative to the article, two phrases matter in this definition: publicly accessible and Internet Protocol. Tellingly, the article uses the expression “parallel Internet”. This likely means that the concept is to connect these servers together while being invisible to today’s public Internet. That is, they want their type of traffic to avoid any equipment/routers that support other Internet traffic. Why? Presumably they need to ensure this traffic is not subject to the type of latency that will occur if the new traffic uses existing Internet gear.
Isn’t this just a description of private networks, such as those based on MPLS and/or QoS techniques? It likely goes beyond this. Note that the examples used in the article included gaming–this implies lots of people who don’t know one another need to exchange traffic. These individuals will have different broadband service providers. One might use Verizon; another might use Level 3. So this parallel network needs to be publicly addressable, just not the same public that is addressable by today’s Internet.
Perhaps this involves a different addressing scheme. Using the telephone network as an example, the phone numbers wouldn’t be 555-5555, it would be XX-XXXXXX. When you dial the use the new addressing scheme, the call is handled by equipment that is entirely separate today’s phone network. Perhaps the second numbering scheme involves a high grade of voice–high definition voice–as today’s circuit switches simply were designed only to support lower quality voice. Every phone call on this network sounds enormously better–but you cannot talk to anyone who is served via a circuit switch anymore.
But perhaps the article goes beyond this. Perhaps they are modifying TCP/IP because the protocol itself isn’t capable of supporting these traffic needs. A new protocol–perhaps with different packet sizes or route forwarding schemes–are better suited for the cloud computing environment.
Can any readers shed light? Ike Elliott, can you help?
Leave a response (2 so far)
Subscribe via RSS
That article is referring to “an IP backbone” constructed to support experimental physics research for the LHC (Large Hadron Collider http://en.wikipedia.org/wiki/Large_Hadron_Collider). One way to think about it is a consortium of research organizations have built their own ISP which only accepts customers who are part of the particle physics research community.
I suspect this backbone has connectivity to the Internet via the many educational IP backbones out there, such as Internet2 in the US. But one likely has to be “on-net” with huge pipes to get full access the capacity the backbone has.
The biggest pipes in this network (multiple OC192’s and long haul 10GigE) interconnect the biggest research facilities that will receive massive amounts of data that is produced by the LHC. This collection of 10’s of thousands of computers is what is referred to as “Cloud Computing” or Grid Computing. Gone are the days of giant supercomputers with hundreds of CPU’s. Now they use 1,000’s of PC’s with gigabit Ethernet between them. Setup a bunch of 10Gig links between 10’s of these Grid computing centers, and you have a “Computing Cloud”.
I believe this network is not really one single backbone, but rather a collection of a few IP backbones. For example, the Starlight research network in the US, the GridPP network in the UK and others.
I think obsolete is not accurate at all. Not big enough and not cheap enough is more accurate. They must just be referring to the fact that these research facilities could not just hook up multiple 10Gig pipes to their local service provider and expect to be able to run them at 100% capacity all day long. Or at least, not in a way that they could afford.