President’s Report - bul.ece.ubc.cabul.ece.ubc.ca/Michael_H.pdfPresident’s Report 2002 09 19...
Transcript of President’s Report - bul.ece.ubc.cabul.ece.ubc.ca/Michael_H.pdfPresident’s Report 2002 09 19...
President’s Report
2002 09 19 Board Report
1
Suite 3200 – SFU Harbour Centre
515 West Hastings Street, Vancouver, BC, Canada V6B 5K3
Phone: 604.822.1348 or 1.800.255.8588 (in BC) Fax: 604.822.9887
www.bc.net - [email protected]
February 9, 2007
UBC ECE Future Networks
Slide 2
The Transit Exchange - a new model
for open, competitive network services
• New and innovative multi-functional networks and services can be easily created using transit exchanges.
• Universities need new networks, and are innovators in this area.
• Transit exchanges are key to these advanced research networks, which are multi-functional.
• TXs help ensure competitive access and encourage network neutrality
• TXs can house disparate wireless technologies
• BCNET has deployed 5 TXs to create its regional network
President’s Report
2002 09 19 Board Report
2
February 9, 2007
UBC ECE Future Networks
Slide 3
Carrier Hotels, MAEs and choice
• End users still connect to the Internet via a single local loop and a service provider backbone network.
• The backbones peer at various Internet exchanges or carrier hotels, the number having grown since 1995.
• This structure is no different than the post-NSFNET 1995 model, where regional networks could choose their backbone provider(s), all of whom were required to peer at the MAEs.
February 9, 2007
UBC ECE Future Networks
Slide 4
Problems with the current model
• It is not clear that this structure will be scalable or even hospitable to new applications and network facilities.
• The peering points become bottlenecks.
• The lock-in to a single provider stifles innovation and keeps overall price points high for next generation network access.
• Issues such as net neutrality loom large with a single provider per user site, as consumers have no choice available.
President’s Report
2002 09 19 Board Report
3
February 9, 2007
UBC ECE Future Networks
Slide 5
Alternatives
• Is there another approach?
• Is there another structure that could better meet the needs of emerging applications?
• Can the overall network structure dovetail with the newly minted functionality being demanded by users?
February 9, 2007
UBC ECE Future Networks
Slide 6
Old and New Networks
Peer-to-peerCentralized
Customizable/flexibleOne size fits all
Whenever you wantReal time, time slots
SymmetricAsymmetric
Content independent from network
Content and network tied
International distributionLimited distribution
Everyone is a producerFew producers
NewOld
President’s Report
2002 09 19 Board Report
4
February 9, 2007
UBC ECE Future Networks
Slide 7
Universities and Networks
Teaching Research Community Service
Experimental Networks
High Performance Networks
The InternetThe Internet
University Mission
February 9, 2007
UBC ECE Future Networks
Slide 8
Connecting communities: The Transit Exchange
• Transit exchanges allow communities to provide FACILITIES without providing SERVICES.
• A community network consists of providers + fibre/wireless facilities.
• Transit links (not a bus ☺☺☺☺) connect this network to the outside world. � An exchange allows the local community easily and economically to
support many transit links.
• The transit exchange is the new central office• Supplier networks interconnect exchanges.
� These suppliers have variability in their connection fabrics.
• The equivalent to free local calling (peering) occurs within exchanges.
• Transit exchanges are the building blocks of new networks.� Paul Vixie: there should be an exchange in every community size 50k,
maybe in the basement of a bank building.
President’s Report
2002 09 19 Board Report
5
February 9, 2007
UBC ECE Future Networks
Slide 9
Have networks really changed?
• Our networks are overlaid on monopoly telcofacilities for the first mile� … well, now it’s a duopoly, and there is a bit of a market in inter-LATA services
� Cellular networks are really the same model, facilities-based with structured carrier interconnect
• Functions have changed, but the basic structure remains the same� Even broadband is a telco term!
• Thesis: next gen networks need next genstructure
February 9, 2007
UBC ECE Future Networks
Slide 10
My needs for any new network
• Solid
• Innovative
• Scope and choice
• Fast
• Cheap
• Simple
President’s Report
2002 09 19 Board Report
6
February 9, 2007
UBC ECE Future Networks
Slide 11
What is the proper structure?
• Structure as (infra)structure� Arrangements of nodes and links
• Lots of thought about function� End2end, p2p, layers
� Routing is a function assuming a generalized structure of nodes and links
• But what about structure?� Is there a structuralist argument for networks rather than black box?
� Is there a philosophy of networks?
• Let function not triumph over structure� There is a delicate interplay
February 9, 2007
UBC ECE Future Networks
Slide 12
Can we create a network marketplace?
• Airports and airlines
• Buses and depots
• Stock and commodity exchanges
• Malls
• What is the equivalent in networks?
President’s Report
2002 09 19 Board Report
7
February 9, 2007
UBC ECE Future Networks
Slide 13
My Network World
• Ubiquitous fibre cabling or wireless connectivity
• FROM: At least 2 independent paths from EVERY location
• TO: independent exchange/interconnect points
• Network exchange points as marketplaces
February 9, 2007
UBC ECE Future Networks
Slide 14
How did current university networks miss the point?
• National, maybe regional at best
• Private
• Poorly interconnected/integrated into community
• Emphasis on bringing the network to campus, rather than the campus to the network
• No first mile emphasis� Rather, long haul networks seem to get the $$$s
President’s Report
2002 09 19 Board Report
8
February 9, 2007
UBC ECE Future Networks
Slide 15
Build the first mile FIRST
• Everything else falls into place easily
• Break free of the duopoly
• Glass is freedom! (as are wireless channels!)
February 9, 2007
UBC ECE Future Networks
Slide 16
First mile fibre – then what?
• Has to connect somewhere and to something� An exchange point and a place to get service
� A network marketplace
• Let’s call these transit exchanges� Different from IXs/MAEs
� End-user oriented rather than supplier
� MAEs/IXs as a solution to the balkanization problem
• Alternative – fibre provider (muni??) as monopoly ISPs
President’s Report
2002 09 19 Board Report
9
February 9, 2007
UBC ECE Future Networks
Slide 17
What is a TX?
• It’s SIMPLE – here are the parts� Patch panel (L1)
� Switch (L2)
� Router (L3)
• Can be done in one rack for a small town
• Connects end nodes to each other and to upstream suppliers
• Allows many-to-many connections
February 9, 2007
UBC ECE Future Networks
Slide 18
TX Components
• Layer 1: fibre and copper patch panels� Allows for cross-connects at the media layer.
� Can have wireless media connectors
• Layer 2: Ethernet switch (10/100/Gb/10GbE)� Supports VLANs, tagging and QoS
� Used to implement multiplexed connections from users to providers.
� Providers only need a single large circuit to service many customers
• Layer 3: peering router� Every user of the TX gets a VLAN to the multilateral peering router.
� User prefixes are exported to all peering participant.
� No other routing is done by the TX operator.
President’s Report
2002 09 19 Board Report
10
February 9, 2007
UBC ECE Future Networks
Slide 19
How it all works …
February 9, 2007
UBC ECE Future Networks
Slide 20
Other TX’s
President’s Report
2002 09 19 Board Report
11
February 9, 2007
UBC ECE Future Networks
Slide 21
Location, location, location
• Optimal spot for both inter-TX service providers and local fibre infrastructure
• Adjacent to cheap colo facilities� Need for expandable real estate
� Do we need zoning for these things?
• We’ve had bad experiences with putting them in city hall
February 9, 2007
UBC ECE Future Networks
Slide 22
How does a TX function?
• Users have fibre to the TX
• Suppliers have fibre to the TX
• TX operator enables cross connects from users to suppliers using L(i)
• Peering router used as a commons� All users connected in a mesh is not scalable, so reduces to O(n)
� But could be used in other ways
� Users (as opposed to suppliers) LIKE to peer with each other – this has VALUE
President’s Report
2002 09 19 Board Report
12
February 9, 2007
UBC ECE Future Networks
Slide 23
Uses of TXs
• Ability to multi-home� Maybe even test multi-path routing techniques� Robustness
• Especially if fibre network is constructed with every node having 2+ independent paths to different TXs
� Building reliable networks out of a set of unreliable providers
• Ability to construct specialized private networks� Use of packet or circuit switching to do so� Users can buy inter-TX links that optimize throughput over
reliability (eg, R&E nets)
• ISPs with NO facilities! Pure brokers� Blended schemes, market makers, IHN� Co-ops or buying clubs can form� VNOs
• Measurement• Multi-TX VPNs
February 9, 2007
UBC ECE Future Networks
Slide 24
Benefits of TXs
• TXs in communities� Encourages competition, driving down prices� Builds technological centre� Can provide a hub to attrach hi-tech business� Allows a platform for application service providers
• TXs as regional hubs� Keep local traffic as regional as possible� Allow remote community ISP (and maybe end users) larger choice
of providers at TX, rather than only the ones who can get to thecommunity.
• Provide incentive for muni fibre networks� Allows city to provide cabling only and staying out of the service
business
• Drive cost of transit down� Transit costs decreased by 30%, capacity quadrupled
President’s Report
2002 09 19 Board Report
13
February 9, 2007
UBC ECE Future Networks
Slide 25
TXs and Transit Costs
Transit Cost and Capacity Trends
0200400600800
10001200140016001800
FY03
04
FY0405
FY0506
FY0607
FY0708
Capacity(Mbps)
Total AnnualCost($1000s)
Unit Cost ($per Mbps)
February 9, 2007
UBC ECE Future Networks
Slide 26
A word about net neutrality
• This is a market condition not a technical one
• Or, it is a regulatory condition, not a technical one
• Protocols/techniques can and will discriminate
• Net neutrality shows that structure has been ignored� Bandaid of regulation
President’s Report
2002 09 19 Board Report
14
February 9, 2007
UBC ECE Future Networks
Slide 27
Questions
• What routing protocols and SW is needed for massive multihoming, peering, and provider switching on the fly?
• What is the right number of TXs in an area?
• Should TXs be aggregated? At what level?� Regional L2 TXs vs close-to-home TXs
� mix and match?
• Who should operate a TX?� Could be privately owned and operated
� InvisibleHand Networks
� Could be independent non-profit (airport authority)
• Should there be competing sets of TXs?
• How can these ideas be tested?
February 9, 2007
UBC ECE Future Networks
Slide 28
BCNET principles
• Building blocks� Use of community dark fibre in local metro networks� Create transit exchanges in local communities where BCNET member institutions (universities) exist
� Connect exchanges with lightpaths� Purchase transit for institutions at each local TX
• Features� Allow ANYTHING – L1, L2, L3� Emphasis on easy local peering and multihoming� Exchanges are open to ALL� R&E networks are simply private networks using community fibre infrastructure and interexchangewavelengths.
� Creates a marketplace per metro area, driving down costs, making interconnects easy
President’s Report
2002 09 19 Board Report
15
February 9, 2007
UBC ECE Future Networks
Slide 29
BCNET – how to build a network
• Acquire local fibre for end nodes� City fibre build issues
• Trend towards city facilities
• RAV line in Vancouver
� Local fibre is expensive unless muni steps in
• Create a TX� Location, location, location
• Attract suppliers� Includes high speed R&E network facilities
• Integrate
February 9, 2007
UBC ECE Future Networks
Slide 30
BCNET Members
BCNET
Core members(Founding)
Higher Education Members
R&E Members
UBC SFU
UVIC TRU
Emily Carr Royal Roads
Okanagan College
UNBC BCIT
NRC
Genome SciResearch Centre
Cancer Research Centre
TRIUMF
CBC
Health Authorities
President’s Report
2002 09 19 Board Report
16
February 9, 2007
UBC ECE Future Networks
Slide 31
BCNET Network Management
Overall Management
UBC24x7 operations
Network Engineering
VANTXUBC managed
PGTXUNBC managed
KELTXUBCO managed
KAMTXTRU managed
VICTXUVIC managed
SURTXSFU managed
February 9, 2007
UBC ECE Future Networks
Slide 32
Western ROADM Project Fibre
President’s Report
2002 09 19 Board Report
17
February 9, 2007
UBC ECE Future Networks
Slide 33
Western ROADM Project Wavelengths
SeattlePNWGP Calgary (Bell)Kamloops (Bell)
VancouverVANTX
VancouverUBC/TRIUMF
KamloopsKAMTX
KelownaKELTX
B7Ch1
B7Ch2 B7Ch2
B7C
h2
B6Ch4
B6Ch1 B6Ch1 B6Ch1
B6Ch2 B6Ch2
B7Ch3
B7C
h1
B6Ch2
B6Ch4
B6Ch2B6Ch2
B6Ch3 B6Ch3 B6Ch3
CalgaryUoC
B7C
h4
B6C
h3
B6C
h4
B7C
h4
B6Ch1
new
MSPP
new
MSPP
new
MSPP
new
MSPP
Victoria (Bell)
B7C
h3
B7Ch3
VictoriaVICTX
existingMSPP
existing
MSPP
try HDXc
foreign
wavelength
(and/or
new MSPP
in Q3 '07)
*only wavelength to be lit immediately*Legend
orange: 10GbE
blue: OC-192B6: CANARIE +NETERA group
B7: BCNET group
move
existing
MSPP
from GT
in Q3 '07
February 9, 2007
UBC ECE Future Networks
Slide 34
Connecting BC to CA*net4
President’s Report
2002 09 19 Board Report
18
February 9, 2007
UBC ECE Future Networks
Slide 35
The Global Research Network
February 9, 2007
UBC ECE Future Networks
Slide 36
Closing thoughts
• New networks need new ways of thinking about infrastructure
• Open access fibre and transit exchanges are key components for next generation networks that enshrine competition as a feature
• BCNET has created 5 TXs across the province
• Next challenges� VLAN and lightpath on-demand management using UCLP or other tools
� Routing issues for massive multihoming
� Wireless handoffs between exchanges (and even multiple providers!)
� True marketplace for users, software selectable