Sign in with
Sign up | Sign in

How The Internet Got Its Hourglass Shape

By - Source: Georgia Tech | B 24 comments

The Internet's curvy figure.

Researchers at Georgia Tech have come up with an hourglass model that consists of six protocol and application layers that originate from a single bubble - IPv4. The scientists understand this model called EvoArch as an evolutionary process that leads to conclusions why some protocols survive and others do not. These new understandings could be used in future developments to evolve the Internet and help develop new uses and better security, the researchers said.

"To avoid the ossification effects we experience today in the network and transport layers of the Internet, architects of the future Internet need to increase the number of protocols in these middle layers, rather than just push these one- or two-protocol layers to a higher level in the architecture," said Constantine Dovrolis, an associate professor in the School of Computer Science at the Georgia Institute of Technology.

According to the model, there is a "plausible explanation" for the Internet's hourglass shape: "At the top, protocols are so specialized and selective in what underlying building blocks they use that they rarely compete with each other. When there is very little competition, the probability of extinction for a protocol is close to zero," the researchers explained.

"In the top layers of the Internet, many new applications and application-specific protocols are created over time, but few things die, causing the top of the hourglass to get wider over time," said Dovrolis.

In the higher layers, a new protocol can compete and replace an incumbent only if they provide very similar services, according to the EvoArch model. An example would be HTTP that largely replaced FTP. At the bottom, protocols serve as "general building blocks and share many products in the layer above." However, no bottom layer protocol can dominate as they are "used in an abundant way", which protects them from being eliminated. The researchers mentioned Ethernet as an example.

The key conclusion of the model is that few powerful and old protocols will be in the in the middle layers that are referred to as "evolutionary kernels", which include the pillars of the Internet and ensure its stability - IPv4, TCP and UDP. While these elements are difficult to replace, there can be a replacement, according to the scientists. The EvoArch model suggests designing protocols "that are largely non-overlapping in terms of services and functionality so that they do not compete with each other." Once the overlapping factor reaches 70 percent of their functions, protocols begin competing with each other.

Quality of the protocol is not always a guarantee for its success: "It is not true that the best protocols always win the competition," noted Dovrolis. "Often, the kernels of the architecture are lower-quality protocols that were created early and with just the right set of connections."

Discuss
Display all 24 comments.
This thread is closed for comments
Top Comments
  • 20 Hide
    juanc , August 24, 2011 4:09 PM
    I sometimes feel I those PhDs are more or less as intelligent as garbage collectors.
  • 16 Hide
    SapienChavez , August 24, 2011 4:29 PM
    juancI sometimes feel I those PhDs are more or less as intelligent as garbage collectors.


    education, like ignorance, has nothing to do with intelligence.

  • 13 Hide
    Benihana , August 24, 2011 5:53 PM
    As they say:

    BS: Bullshit
    MS: More shit
    PhD: Piled higher and deeper
Other Comments
  • 20 Hide
    juanc , August 24, 2011 4:09 PM
    I sometimes feel I those PhDs are more or less as intelligent as garbage collectors.
  • 10 Hide
    acadia11 , August 24, 2011 4:15 PM
    The question if it ain't broke, why fix it, TCP functions fine as the backbone of the internet why replace it. Unix/and it's derivatives Linux function great as OS model and have for some 40 years, again, why replace it. The more things change the more they stay the same. For example, cloud computing, it's the Mainframe model outside of the closed eco-system that was most mainframe architecture, i.e. the thin /dumb client ... connecting to the heavy duty cloud (main frame) which has your app storage, and data storage.

    It's now just been integrated to "Cloud" that is internet based, as opposed to a closed architecture.

    The wheel has worked for some 60,000 years, its same basic protocol has not needed to be changed, with that said. Unless, there is a need to replace it, why do so, that's the point the researchers seem to miss.
  • 7 Hide
    Anonymous , August 24, 2011 4:21 PM
    That for sure goes in my list of the 10 most stupid articles I've read in my life; by reading it I've lowered my IQ some points.
  • 16 Hide
    SapienChavez , August 24, 2011 4:29 PM
    juancI sometimes feel I those PhDs are more or less as intelligent as garbage collectors.


    education, like ignorance, has nothing to do with intelligence.

  • 2 Hide
    COLGeek , August 24, 2011 4:39 PM
    To quote the great Homer Simpson...."Doh!!!"

    I wonder how many dissertations were based on this study?
  • 1 Hide
    cypeq , August 24, 2011 5:13 PM
    This picture is missing at least 40 arrows but I get the point to make it simplistic but I just don't know why internet construction never was and never will be so simple.
  • 3 Hide
    nicodemus_mm , August 24, 2011 5:17 PM
    acadia11The question if it ain't broke, why fix it, TCP functions fine as the backbone of the internet why replace it. Unix/and it's derivatives Linux function great as OS model and have for some 40 years, again, why replace it. The more things change the more they stay the same. For example, cloud computing, it's the Mainframe model outside of the closed eco-system that was most mainframe architecture, i.e. the thin /dumb client ... connecting to the heavy duty cloud (main frame) which has your app storage, and data storage.It's now just been integrated to "Cloud" that is internet based, as opposed to a closed architecture. The wheel has worked for some 60,000 years, its same basic protocol has not needed to be changed, with that said. Unless, there is a need to replace it, why do so, that's the point the researchers seem to miss.


    A bicycle works fine for travel, but I'm not going to use it to go from Maine to California. The wheel also isn't made of stone now. The idea that because something "works" we shouldn't research improvements or replacements is ludicrous. Sometimes change brings about innovations in efficiency, reliability and new features to improve the lives of those using the product. That's how breakthroughs are discovered. Most of these are points that you seemed to miss.

    ~ Nicodemus
  • 7 Hide
    theoldgrumpybear , August 24, 2011 5:29 PM
    And how many "new" patents will be filed with this model as basis?
  • 13 Hide
    Benihana , August 24, 2011 5:53 PM
    As they say:

    BS: Bullshit
    MS: More shit
    PhD: Piled higher and deeper
  • 2 Hide
    RazberyBandit , August 24, 2011 6:06 PM
    This would have been great 10 to 20 years ago when IPv4 was still practical, but it's limitation of 4.3 billion unique addresses doesn't work today. IPv4's days on the Internet are numbered. The population and technology booms since it's adoption have significantly reduced it's expected lifespan. It'll still work great for closed networks, but it's no longer sustainable when it comes to outside communications.

    These guys should refocus themselves towards building a similar model for the future based on IPv6.
  • -2 Hide
    Anonymous , August 24, 2011 6:42 PM
    That article seems like it's a fairly trivial observation, followed by a conclusion which is supported by neither the observation nor by common wisdom.

    I personally cannot think of a single reason for having multiple, fast evolving, narrow application protocols competing for the position held by IP. Just see how much trouble it has been to replace IPv4 with IPv6 despite urgent reasons to do so. Imagine having multiple organizations deplying regular changes at that level - I cannot fathom how anyone remotely ascociated with computers could thing that would enhance security in any meaningfull way, and the does nothing to prove it either. Yet somehow it arrives at that conclusion.
  • 4 Hide
    palladin9479 , August 24, 2011 7:42 PM
    Quote:
    These guys should refocus themselves towards building a similar model for the future based on IPv6.


    While I'd admit that we need to evolve past IPv4, the guys behind IPv6 made some serious mistakes when they designed it. They designed a protocol for a "perfect" world and not a real one and have resisted efforts of the business world to get them to change it. One is that 128 bit's is entirely to freaking long for a unique address, heck 64 is too long, but somewhat doable. Next is the way of encoding address's only works for router guys, for everyone else it's not a practical method. Then they try to force you to use IPSEC, something that has broken compatibility depending on how the vender implements it. Products from different venders trying to use IPSEC between them run a 50/50 chance of it just not working. It's so bad that NSA has designed their own HAPIE specification that use's RECIPe (Remote Encryptor Configuration Information Protocol) to enable and setup the tunnels (it's a heavily modified IPSEC implementation). And to final straw is that they refuse to support any implementation of NAPT (what was have in the IPv4 world), which immediately stops many companies from switching over. It's so bad that a Chinese college student went out and built a NAPT66 module for Linux.

    China has already changed over to IPv6 and guess what, it didn't do the "nice" thing and give everyone a full /64 range like the IPv6 guys said they would. China also used IPv6 to track, register and catalog every single network device in their country. And due to the forced "end to end" model it allows China's government to track exactly which websites and internet address's that every single Chinese citizen has been to.

    So yeah, we're stuck with IPv4 until something better is made, possible IPv7/8 as IPv6 in it's current implementation isn't an answer.
  • 0 Hide
    WyomingKnott , August 24, 2011 8:05 PM
    juancI sometimes feel I those PhDs are more or less as intelligent as garbage collectors.

    Nah, the Garbage Collector in the Java implementation that I use is pretty clever.
  • 0 Hide
    WyomingKnott , August 24, 2011 8:06 PM
    nicodemus_mmA bicycle works fine for travel, but I'm not going to use it to go from Maine to California.

    Why not? I went from Portland, Oregon to Pueblo, Colorado by bike the summer that I was seventeen; it was a great experience.
  • 2 Hide
    RazberyBandit , August 24, 2011 8:37 PM
    Quote:
    Why not? I went from Portland, Oregon to Pueblo, Colorado by bike the summer that I was seventeen; it was a great experience.

    The fact that you're only able to cite one such example speaks volumes in regards to long-distance cycling feasibility...
  • 2 Hide
    tacoslave , August 24, 2011 9:33 PM
    WyomingKnottWhy not? I went from Portland, Oregon to Pueblo, Colorado by bike the summer that I was seventeen; it was a great experience.

    time
  • 0 Hide
    boiler1990 , August 24, 2011 10:41 PM
    Georgia Tech obviously doesn't know that the Internet is a series of tubes...
  • 2 Hide
    back_by_demand , August 24, 2011 11:10 PM
    theoldgrumpybearAnd how many "new" patents will be filed with this model as basis?

    Apple is working on it as we speak
  • 0 Hide
    eddieroolz , August 24, 2011 11:20 PM
    Soon we'll have a widening of the hourglass with IPv6.
  • 0 Hide
    Undeadhunter , August 25, 2011 7:30 AM
    That's just the freaking OSI model ... http://en.wikipedia.org/wiki/OSI_model

    How is this new?
Display more comments