Future Internet Vision, Workshop Scope & Objectives
Dipankar Raychaudhuri: This is an NSF initiative aimed at building new experimental infrastructure for the country. It is aimed at getting multiple groups together to identify important research priorities; We have had some discussions on mailing list regarding what is required of an experimental infrastructure.
The agenda outlines three groups. Tomorrow we will have breakout sessions of the groups where there will be detailed discussions and recommendations will be made. Later tomorrow, we will meet again for final decisions and we will leave the workshop with some writing to do.
Opening speech by Dave D. Clark – MIT CSAIL
Title: A new Architecture for an Internet.
The Ideas presented in this workshop is an ambitious idea inside NSF. It is time to build a new Internet. This is the basis for this concentrated effort behind the infrastructure experiment.
Why a new Internet ?
In 10 – 15 yrs need an Internet with different characteristics than what is supported today. Today’s internet is focused on data forwarding. A different Internet should be envisioned. The current approach to research directions to improve the Internet, for example with the IETF is similar to a “Random Walk” away the current Internet. Its short term and lacks a common goal. We need to free the mind from current model and ask the question :If you would do it now, what would you do ? Then keep good ideas and throw out bad ones.
The Networking community needs to again start saying – “we can change the world”; New internet will motivate the whole thing of experimental infrastructure. We could change the world. Why might we want a different Internet in 10- 15 years ?
- Better Security: compelling justification. Today, there is no framework for security – Internet comprises of various security pieces – firewalls etc; Networks need to be designed to focus on availability and robustness. Think about what is the network’s responsibility to protect end nodes.
- Adjustment for Industry structure and economics.
- Network management: P2P link covers tremendous amount of physical complexity – SONETs, Ethernet etc. Abstraction do not fail, but physical links do; There is no architecture to represent the physical links today. As a result, the Internet is difficult to configure and get working for most people.
- Need better ways to incorporate new network technologies such as new wireless networks, fiber optic networks etc.
- Need to service both public and priv. nws.
- Internet should serve in times of crisis. Is the Internet meeting the social needs today ? Need to reconceptualize social responsibilities such as 911 etc and evaluate what is important for the Internet.
Support new computing:
The Internet should be that thing that puts cool tech together. Look at what computing will look like in 10 years, do not start with networks.
It will be a decade of cheap ubiquitous low power. Not performance. The $1 device – the 10 cent network. There used to be a time when many people shared a computer – time sharing and user security were important research issues then. Now each person has a computer. Now need to “get unstuck from 1”; Its already happening - many computers for 1 person. Need to assess what the Internet should do in that world.
The decade of the application; what do we (networks) do to support applications ? Need to ask how applications are built just as much as we think about building packet carriers.
If processing is Everywhere:
New management paradigms ? how do systems organize and how do parts find each other ? bob and alice’s computers found each other automatically – Identify a crucial role of physical location. Does the network play a role in this ? new security issues – what to do when my computer accidentally goes on a date with a bad one ? Is infrastructure open or integrated ? can I go to radio shack and buy a sensor that attaches to an open network ? Is there going to be an open infrastructure for the wireless internet ?
The reality of applications:
They are not “end-to-end”! think from apps in the context of the network. There are lot of intermediate nodes - email does not go directly from my computer to yours. Instead it goes from mine to yours via SMTP servers, routers etc.
Applications are location-aware. Eg. Acamai – if nearness did not matter they wouldn’t exist. They capilized on what ISP’s offered for free – web caching.
They are the source of many security problems.
They are the space of tussle – turbulence that is shaping the Internet.
They are the driver of value – all packets generated by applications.
Bits to Rethink
Packets and multiplexing; Why don’t packets have license plates ?!
Addressing location and identity;
Layering ? should there be something like TCP in the future Internet ?
Application support – people building overlays. Is this the right approach ?
Some general thoughts
Excitement will be at the edge. Workshop is centered where excitement is. New computing, new networks and new applications.
Tell people the story they want to hear:
Better security; innovation and bigger pies.
Usability: in a computing workshop get others to write research agendas – physicists said, you’re pretty fast, can you make it easier to use? This applies for networks too.
Don’t forget – the Internet is global. Take Internet away from US to save it. UN has a diplomat assigned to manage governance of internet.
- Comment: Internet explosion came from email, web browsing etc; branding the Internet is happening today.
- Deb. Estrin: going specifically to human to human communication is a narrow outlook
- Rich Howard: need to introduce voice – take GSM networks into consideration
4. Jay Lepreau: Application focus – systems guys always focus on this. Terrible at it even now. Need to broaden playing field for it.
Title: CIRI (Clean Slate Internet Reinvention Initiative) and GENI (Global Environment for Network Investigation)
Envision a global network infrastructure that is lot more responsible, supports future sensor technologies etc. The Internet today is not ready for future roles.
Fundamental limitations are in security, scaling to be inclusive of diverse devices; several planning activities required from this workshop.
The NSF workshop “Overcoming barriers to distruptive innovations in Networking” concluded that the problem was with the current Internet. Build on foundation that is not correct.
These problems are opportunities for a technology push. (sensors, data mining, new algorithms etc); There is also significant commercial interest in health care, etc.
CISE portfolio – a perspective
Good news – positive response from community to reinventing Internet. This workshop needs to identify goals and requirements. The research focus needs to be on Reinvention of the Internet.
Scope of GENI – everything from edge networks to applications. Vacuum between existing network testbeds to finally deployed commercial networks. GENI is not just another testbed. It needs to be a meta testbed – a virtual testbed. Should support many architectures specific network testbeds. Should exploit advantages of diverse network technologies.
Expected GENI deliverables – new class of a. network platforms – routers, switches etc.
GENI targets funding of $300M over 5 years. Systems – prototypes to commercial grade infra.., experimentation. ; time frame – FY08 is earliest reasonable.
The Reinvention initiative and Experimental Infrastructure effort fit together; Need workshop input to support both. Meta planning grants – purpose- articulate research agenda. Robust experimental testbeds.
Need a compeling research agenda:
Go beyond current internet; be comprehensive; be synergistic – this has to be much more focused than typical ones- decide experimental Infrastructures. Need exp. evaluation and deployment plans; needs to include a consensus bldg process.
Agenda audience and goals
Need to understand how will lead to growth in the economy
Network funding is relatively healthy with 15% acceptance rate.
Without compeling research agenda funding will go down. Acceptance rate will go down to maybe ~5% as in other programs; Need to propel in the direction of better proposals. Non exciting programs pretty hard to defend.
Exciting research agenda will help keep existing generous networking budget. More discretionary funds to networking.
There is thus a much greater urgency to innovate beyond Internet. NSF ready to commit serious resources.
1. 2-3 years ago NSF funded couple of approaches. How are those goals diff from current ones.
Answer: for wireless – reports typically tended to list of diverse research projects rather than a converging set; They were loosely connected, minor improvements to the Internet;
Guru: Based on awards made, one may get an idea of NSF direction. Reacting to input from community.
2. In the past – a large initiative such as super computing ; does this initiative require participation from other communities?
Eventually an opportunity to support science and engg communities; But this workshop is meant to investigate building Experimental Infrastructures for Networking and systems research. Some of those other communities will be able to participate here through natural use of infrastructure. I&E; infrasturctier for networking and systems research; focus on this rather than other communities.
3. What value do users get to opt-in ? typically bandwidth. Need to think what value we can provide. Offer something say security in wired internet.
4. Effort to explain to the ISPs :currently they’re torpedoing a lot of efforts.
Ans: hopefully as a community we can convince them. Input: from industry perspective – NSF is a face to government involment; need more
Title: Wireless, Mobile, Sensor Technologies and the Future Internet
Wireless is the key driver for the future Internet
In the future, all cell phones gradually migrate to the Internet. 500 million in-connected PCs in 2005; new wireless infra networks – mesh, wimax; radio router as n/w tech. other types of devices not envisioned by current Internet.
In the future M2M (machine to machine) may be dominant form of communication.. addressing needs to be revisited.
Infrastructure nodes in wireless; - security, achieve pervasive networks the.
Broadband was the big topic in 90s. now includes functionalities such as security; power efficiency and other computer constraints are important.
Future wireless needs to move away from today’s vertical stack. Majority of wireless devices are considered to be off the Internet. They go through gateways via PSTN. Future of Internet - very large edge networks, In home broadband networks, Sensor nets etc; Wireless architecture can be treated as a separate piece of the Internet – cut off as a subnet; This argument is becoming weaker as wireless networks form larger and larger pieces of the internet.
We have a once in a lifetime opportunity to design a more integrated architecture.
Designing a new Internet Architecture:
TCP/IP evolution or revolution ?evolve Ipv6 or Ipv8 ? Or do we need distruptive innovations;
Caveats: The Internet architecture we design should be different from the bulk of previously done utopian work in networking.
Experimental infrastructure: Develop pragmatic systems.
An Open access link model is preferred. PlanetLab uses time sharing and CPU slicing approach. It is different in emuLab and ORBIT where the entire testbed is utilized by a single experiment at a time. This approach is preferable with wireless systems, particularly with the interference issues involved. There is also a 3 rd approach – the protocol labeling approach used in Active Networks.
High level goals; Some questions for group (please see the slides); go to website to look at the questions and you may also add more.
Please see slides for full details of the talk