OSCnet/OARnet Engineering Update

Download Report

Transcript OSCnet/OARnet Engineering Update

TFN/OARnet Engineering
update
OARtech April 12, 2006
Paul Schopis
Last Mile Update
• Hebrew Union College
• Delayed due to Q-in-Q issue with TWC
• Southern State Community College
• Rehoming lines for better utilization
• North Central
• New T1
• John Glenn
• Rehoming to Cleveland to close POP
• Oberlin
• Serial card delayed due to slow capital turn around
Support Center
•
•
•
•
•
Two tiered
Facilitate training
Facilitate better division of labor
Facilitate faster turn around
3
Partners
• Some K-12 ITC sites are on TFN
– Toledo Public
– SCOCA
• Public Broadcasting
– WVIZ Cleveland
– Waiting for additional middle mile for
additional sites
New Directions
• Merit direct interconnect
– Peering with Merit Orano and NYSERNET
– Potential push eastward with MERIT
– Chicago access
New Directions
• Chicago
–
–
–
–
–
–
–
Joint venture with OSU
Take advantage or fiber rings
OMNIpop
NLR
Future Abilene?? (Non-merger hedge)
Commodity
Additional peering
• SLR, BORISNET
pDNA
Programmable Dynamic
Network Architecture
Paul Schopis, Chris Heerman &
Tsege Beyene
The Problem
There are numerous programs under development to
revolutionize the current “physical” internet. Most programs
attempt to reach their goals by a virtualization of resources or
putting in the end users hands access to resource allocation.
Some of the leading programs are Planet Lab, Hopi, UCLP
and Dragon. Most of these concentrate on subset of issues
relating in one way or another with getting dedicated resources
on demand or semi-on-demand. All require a preconfigured
node or device to allow use of a resource whether physical or
logical. But what if we extended this notion to the control plane?
What we propose here specifically is a provisional control plane
mechanism for total resource creation and control.
Planet Lab
• Xen virtual machine
• Allow researchers to research protocols etc
• Uses real routing information acquired from the
commodity internet to present a virtual environment
as a “real” simulation
• Gives fairly serious resources to researchers e.g. a
researchers playground
• Overlay network
• VINI
Hopi
• Hybrid Optical Packet Infrastructure
• Leverages network layers to construct a
virtual “Light Path”
• Deployed but still very conceptual
• Working with Dragon on GMPLS etc.
UCLP
• User Controlled Light Path
• Gives end user access to optical
resources
• Ethernet in Sonet Framing
Dragon
•
•
•
•
Uses GMPLS - current sets up VLANs
Allows dynamic lambda switching
Resource Scheduling
Virtual Lambda Switching Router
(VLSR)
• Network Aware Resource Broker
(NARB)
GENI
• Based on Planet Lab
• Employs notion of virtual router
• Highly specialized case; private domain
OARnet/TFN
• Uses MPLS for recovery & partitioning
• Uses Logical Routers in production
• Juniper LR is others VR
Proposal
• What if….
– Leverage Planet Lab virtualization and
extended overlay notion to…
– HOPI like light paths and…
– used Dragon as part of the subset for
signaling
– Finally, what if we could dynamically
provision the control plane much like the
resources described?
GMPLS
LFIB
Architectural BLDG Blocks
Control Plane/Service Domain
Network Entity
Intra-Domain
Inter Domain
Forwarding
Component
OUNI
Manager
Optical
switch
L2UNI
Manager
Layer 2
switch
Management
Plane/Security
Net Monitor
L3UNI
Manager
Layer 3
The Problem
• The $50,000 question is how is resource discovery,
binding and provisioning performed in global and
scalable way?
• Can we use and slightly alter mechanisms that
already exist?
• In other words can we use an existing infrastructure
as an out-of-band control channel to facilitate creating
new infrastructures?
Assumptions
• Facilities (POPS) that support new infrastructure
deployment generally we be located at facilities that
support the current internet
• Multicast attributes can be leveraged for resource
discovery and bootstrap provisioning
• Assuming that both prior assumptions are true we can
use the existing internet as an out-band-control
channel
• The advantage is that it already exists and is
ubiquitous in the research and education community
Proposed pDNA Site
Servers (real or virtual)
To
Internet
MBGP
Router
Layer 2
Switch
Router Attached to
Standard Internet
Optical
Resources
Layer 2
Resources
Resource
Manager
Admission
Control Server/
Security &
Authentication
Network Monitor
Layer 3
Resources
Functional Overview
RM
Internet
Router
RM
Router
Multicast Resource advertisement
Functional Overview
Advertisement
Heard & Stored
as point of interest
RM
Internet
Router
RM
Router
Multicast Resource advertisement
Functional Overview
Bootstrap Exchange
Request
RM
Router
Internet
RM
Router
Functional Overview
Authentication and Authorization
Challenge
RM
Internet
Router
RM
Router
If {yes proceed}
Elseif {no
Goto next point
of interest}
Else {(fail) respond
to user exit}
Functional Overview
Resource Exchange
RM
Router
Internet
RM
Router
Functional Overview
Resource Request
(protocol X(RSVP?))
RM
Router
Internet
RM
Router
Functional Overview
Resource Response
RM
Internet
Router
RM
Router
If yes proceed
Elseif no
Goto next point
of interest
Else fail respond
to user exit
Functional Overview
RM
Internet
RM
Router
Switch
Router
Switch
Server
Server
Code transfer
can be Modules
Support IP, GMPLS
or Experimental Protocols
Etc.
Functional Overview
RM
Internet
RM
Router
Switch
Router
Switch
Server
Server
Routing
or Protocol
Engine
New Infrastructure
Optical, Layer 2,
Layer 3 or Overlay
Routing
or Protocol
Engine
Functional Overview
RM
Internet
RM
Router
Switch
Router
Switch
Server
Server
Routing
or Protocol
Engine
New Infrastructure
Optical, Layer 2,
Layer 3 or Overlay
New Dedicated Domain
Routing
or Protocol
Engine
Conclusion
• The afore mentioned process does not
preclude an iterative and nested process
• As such a process could be used widely and
global paths
• For example, Servers could be allocated as
LSR routing engines on link by link basis to
construct an LSP if needed.
• This might be a way to address the interdomain resource problem
Questions?