from arpanet to lambdagrid: the 1980s eruption

28 September, 2005 Slide 1

From ARPANET to LambdaGrid: The 1980s Eruption

Dennis Jennings

NSF Program Director for Networking (1985/86)

ex University College Dublin



• Background / Context– The “Internet”

• concept invented; research phase completed.• no large scale demonstration.

– US (Academic) Research the driver



Start of the Decade



• 1980– Telecomms: Low Cost, Low regulation– Campus Networks: None– Networks: ARPANET, MFENET, Usenet– TCP/IP Implementations: BBN/ARPANET– Routers: BBN/ARPANET– Network Protocols: Proprietary - Decnet, SNA,

VNET, Misc.– ARPANET success inspired – but access limited – PC / Workstations – yet to come



1980 to 1984



• 1980 – 1984– Telecomms: ***

• More competition / more deregulation• T1, T3 Circuits available and cost effective • Falling costs

– Campus Networks:• Ad-hoc departmental• All standards – Ethernet (1980)• Few ubiquitous networks



• 1980 – 1984– Agency Networks:

• MFEnet – DECNET – 1976/1980 (ESNET 1985)• ARPANET – (TCP/IP 1983) ***• SPAN – DECNET – 1981

– Community Networks:• Usenet• BITNET – VNET – 1981 • CSnet – TCP/IP – 1981 (+ARPANET) ***



• 1980 – 1984– TCP/IP (DARPA Protocol Suite) Implementations:

• BSD 4.2 Unix – 1983 (DARPA) ***• ARPANET – TCP/IP (1983)• VAX: Wollongong (1984?)• IBM: Wisconsin (1984?)

• Cray: None



• 1980 – 1984– Main Protocols:

• DECNET – VAXen• SNA – Mainframe number crunchers

• TCP/IP – Unix Workstations – Sun *** – Routers:

• BBN• Cisco/ Proteon/ Ungermann Bass (1984) • Ad-hoc



1984



• 1984– Recognition:

• Networking important.• CSnet success• NSF’s Advanced Scientific Computing Initiative (OASC)

– network required

– Budget:• Allocated for supercomputer network (NSF – ~10%)

– Program Director• Appointment made – to start 1 January 1985



1985



• 1985 – NSFNET Key Decisions– NTAG

• To appoint a technical advisory group to advise (and, more generally, to build consensus among the US academic research community)

– General Purpose Network:• to develop a general purpose national network for all

(science and engineering) research, not just a specialised network for supercomputer users, in order to achieve the mission goals - i.e. to provide all of the science and engineering research community with access to supercomputing power (and by extension to all researchers/academics)



• 1985 – NSFNET Key Decisions– An Internet:

• to develop a network of networks – an Internet – rather that a dedicated single supercomputer network.

– TCP/IP:• to adopt the TCP/IP (DARPA) protocol suite – the

Internet protocols. • to mandate that the TCP/IP protocol suite be adopted (in

due course) by all existing supercomputer networks funded by the NSF’s Advanced Scientific Computing programme.



• 1985 – NSFNET Key Decisions– Architecture:

• to build a three-tier network – based on:– campus LANs, – Regional, Supercomputer, and community networks

(e.g. CSNET, BITNET), ARPANET, and – a national backbone network - linking the national

supercomputer centres/networks (and through them the regional and campus networks).



• 1985 – NSFNET Key Decisions– To fund mid level networks

• the supercomputer centre networks

- the JvNC network, provided it was TCP/IP; *

- the SDSC network, provided it migrated to TCP/IP in due course

• the start-up of State and Regional networks (e.g. NYSERNET, SURANET, BARRNET, etc.) and to encourage local funding or these networks.



• 1985 – NSFNET Key Decisions– (To fund mid level networks)

• the start-up of a number of “high speed” (T1 – 1.5 Mbit) networks.

• the CSnet Cypress project• the BITNET TCP/IP project



• 1985 – NSFNET Key Decisions– Not to fund campus networks

• not to fund the campus networks, but to leave to the campuses the responsibility for the delivery of networking to the end users (Campus “Gateways” were funded).

– To fund the NSFNET backbone• 17 September 1985



• 1985 – NSFNET Key Decisions– Routers

• to use an interim router solution (Dave Mill’s Fuzzball) for the NSFNET backbone (to permit the industry time to develop routers that met the standard router specification).

• to develop a specification for a standard router (RFC985).



• 1985 – NSFNET Key Decisions - continuedWhile all of these decisions may seem reasonably obvious with hindsight, none of these decisions was seen by everyone involved as obviously necessary or correct at the time, and all of the decisions were controversial, hotly debated, and contested.



1986



• 1986 – NSFNET Early Implementations– Backbone (56Kb)– Supercomputer Centre Networks (56/T1)– Regionals (56Kb) (BARRNET at T1)– ARPANET expansion (slow)– Csnet Cypress Project– BITNET – TCP/IP Project– Campus Networks



1987 to 1989



• 1987 – 1989: Rapid Growth– NSFNET T1 Backbone– Professional Management (Merit/ IBM/ MCI)– Regionals – growth and interconnection –

professional management – privatisation – commercialisation ….

– Campus networks– NSFNET T3 Backbone planned– Everyone wanted to be on the network / Internet– US FCCSET coordination and planning



• 1987 – 1989: The Engineering Phase– TCP/IP implementations– Routing development– Performance enhancement– Tools for information discovery …..



• 1987 – 1989: The Commercial Phase– Commercial router development– Commercial Networks

• UUNET (Commercial – 1989)• PSInet (1989)

– Everyone wanted to be on the network / Internet



… and By the End of the Decade



• 1990– The US National Research Network was in place– The Internet was in place– The Commercial Internet was beginning– The stage was set - for the next Eruption……



……………………..

from arpanet to lambdagrid: the 1980s eruption

Documents

arpanet tcpip

s eruption1984from arpanet

s eruption1985from arpanet

arpanet success

decadefrom arpanet

s eruption1984recognition

s eruption1980telecomms

s eruptionstart