Transcript 20020506-CENIC-StArnaud

CA*net 4
Open Grid Services for
Management of Optical
Networks
CENIC Workshop
May 6, 2002
[email protected]
Many e-Science Lambda Grids
ALMA
LHC
Sloan Digital Sky
Survey
ATLAS
Current Optical Internets
ISP
AS 4
UNI
Carrier controls and
manages edge devices
AS 1
Optical VPN
NNI
AS 1
UNI
UNI
Customer
AS 5
AS 3
NNI
AS 2
Big Carrier Optical Cloud using GMPLS
or ASON for management of wavelengths
for provisioning, restoral and protection
The Problem
 Current optical services are based on GMPLS, ASON or O-UNI
 Current optical services are “edge to edge” within a single carrier
cloud
 Any changes to customer optical VPN in terms of bandwidth or topology
requires release of current VPN and establishment of new optical VPN
 Customer cannot make topology or bandwidth changes within their own
VPN, or cross connect to another VPN within the cloud
 No inter-domain optical routing protocol
 Customer cannot set up an end to end wavelength across multiple domains
 No optical services for end2end light path across the campus/enterprise
and across the carrier cloud
 All current optical services are based on a client –server model
 No ability to exchange wavelengths and services on a peer to peer basis
 University/Research community comes from the Internet world where
services and networks are offered on a peer to peer basis
 BGP multi-homing, mail servers, web, etc
 Peer to peer has enabled the powerful Internet end to end principle
Customer Control of the Network?
 There is a growing trend for many schools, universities
and businesses to control and manage their own dark
fiber
 Can we extend this concept so that they can also own and
manage their own wavelengths?
 Will “empowering” customers to control and manage
their own networks result in new applications and
services similar to how the PC empowered users to
develop new computing applications?
CA*net 4 Architecture Principle
 End user control of the network
 A network of point to point lightpaths where individual users can
manage, control and route their own lightpaths and cross
connects
 Initial users expected to be high end grid applications
 Web service architecture for customer management of lightpaths
and cross connets
 Uses OGSA or WSDL, UDDI, JXTA, J2EE, etc for optical management
 Network becomes another Grid service
 Lightpath owners can recursively sub partition their lightpaths
and cross connects and give control to other entities
 Lightpaths become “objects” complete with polymorphism,
inheritance, classes, etc – Object Oriented Networking
Networks are like Grids
 Management of inter-domain networks face the
same challenges as management of Grids
 Grids allow shared usage of resources in
different management domains
 But more importantly (with web services
architecture) allow for shared and distributed
use of processes (in addition to resources)
Advantage of Web Services
 Elimination of the protocol standards process!!
 Functionality of a service is defined in WSDL as opposed to
being “locked” into a standard
 WSDL and SOAP make it easy to define new services and
functionality without going through complex standards approval
process
 WSDL and SOAP can be bound to traditional API calls e.g. OUNI to provide open interface to users
 Web services will be the foundation of Grids and distributed
computing
 Allows for optical networks to be just another seamless
service in the grid toolbox
Wavelength Assignment
Carrier A
12
10
University
Regional Network D
3
13
15
AS 1
2
4
AS 2
5
7
Regional Network C
LightPath switch
AS 5
14
9
AS 6
8
Carrier B
6
University
STAR LIGHT
router
CA*net 4 Architecture
CANARIE
GigaPOP
ORAN DWDM
Carrier DWDM
Edmonton
Saskatoon
St. John’s
Calgary Regina
Quebec
Winnipeg
Charlottetown
Thunder Bay
Montreal
Ottawa
Victoria
Vancouver
Fredericton
Halifax
Boston
Seattle
Chicago
CA*net 4 node)
Possible future CA*net 4 node
New York
Toronto
Windsor
Initial Version of Mini-IX
External Proxy Server
Subtended GbE to local GigaPOP
OC192 Eastbound
OC192 Westbound
Customer A and
sub- partition
Grooming agents
X
Customer A
Proxy Server
Standard CLI or TL1 interface
Customer A signaling plane
X
OSPF
X
OBGP
X
Customer B signaling plane
Customer C signaling plane
Customer B
X
OBGP
Customer C
Switch Agents
CA*net 4 Proxy Server
Signal Control Plane Agents
e2e LightPaths
 High end users (e.g Grids) can apply for their own end 2 end light path
 Requirements:
 Must demonstrate real need for required bandwidth
 Must implement high performance kernels and IO process to take
advantage of bandwidth
 Must resolve bandwidth bottle necks across campus and local loop
(funding may be available for this)
 If demand is greater than supply then CANARIE may prioritize or
implement charges
 Examples (illustrative purposes only):
 OC48 TRIUMF to CERN
 CWDM across UBC campus and local loop to CA*net 4
 OC48 light path to Chicago to connect to CERN
 WestGrid Interconnection
 SNO Interconnection to HPCVL
OBGP
 Proposed new “web services” protocol to allow customer
owned wavelengths to interconnect to each other at an optical
switch
 Optical switch is in effect a mini-IX
 Use establishment of BGP or peers for process to establish
light path cross connects
 Allows customer to maintain routing policy on cross connect
 Traditional BGP gives no indication of route congestion or
QoS, but with DWDM wave lengths edge router will have a
simple QoS path of guaranteed bandwidth
 OBGP can use “optical” AS path to concatenate wavelengths
across multiple AS to have continuous QoS path
OBGP transit
AS1
AS2
AS3
AS4
AS5
Lightpath Agent
OBGP Agent
Object
Name
Server
Connection Oriented Networking?
 Connection oriented networking is not inherently bad
 Connection oriented networks got a bad rap because of telco
business model and ATM
 Big advantage of Internet was not so much connectionless as the
end 2 end principle
 Because users had control of packets they could develop applications
without getting permission
 Services are also peer to peer and not client server
 It is essential for QoS applications or high volume data transfers
 Two approaches:
 Establish end2end guaranteed bandwidth path over prioritized packets
 Use underlying connection oriented circuits
 Can we extend the Internet advantages to connection circuits?