Transcript Document
Hardware Accelerated
Signaling and its
Application in Fast
Network Restoration
Prof.Veeraraghavan
Prof.Karri
Haobo Wang: [email protected]
Background
What’s signaling?
Signaling
protocols are used in connection-oriented
networks to set up and tear down connections.
Primarily implemented in software
Complexity
and the requirement for flexibility
What we pay is the performance
Hardware accelerated signaling
Why?
100x-1000x speedup
How? FPGA+Hardware/software Codesign
Which is better?
“Thin” hardware signaling engine
Simple
signaling addressed to single application
Optical Circuit Signaling Protocol – OCSP
Simple, for SONET network only
Hardware accelerated signaling engine
“Panacea” signaling – RSVP-TE for GMPLS
Complex but covers all connection-oriented networks
Hardware/Software
Codesign
Time-Critical functions->hardware
Non-Time-Critical functions->software
Network and node view
Architecture of Prototype Board
To Host
PCI Bus
FPGA (Xilinx XC2V40)
PCI Bridge
O/M
(HFCT
53D5)
SERDES
(HDMP
-1636)
GbE
MAC
(L8104)
Hardware
Signaling
Accelerator
TCAM
(MT75W8Y136H)
SRAM
(MT55L64L36P1)
Switch Fabric
(VSC9182)
On-Board Bus
Hardware Signaling Accelerator
GbE Interface
RAM
(Message Buffer)
PCI
Bridge
Object Dispatcher/
Assembler
TCAM
Interface
Register Bank
OP*
* Object Processor
OP
OP
SRAM
Interface
Applications of Hardware signaling
High throughput
Support large scaled core switches – TCP switching
Low call setup delay
Fast
restoration after network fails
Illinois, May 1988, a fire at a local exchange center
caused the loss of service to 35,000 residential
telephones
Today, data traffic prevails, more sensitive to loss
One single fiber may carry T(era)bps of user traffic!
Network survivability is essential
Trade-off of different recovery schemes
Resource utilization
Resource reserved in advance
Slow
Good
Restoration
react to failure dynamically
Mesh networks
Path switching
LAPS or SHR
Line switching
Bad
Linearly/Ring topology
Recovery delay
Protection
Fast
The failure is addressed
locally, point-to-point
The failure is addressed
globally, end-to-end
We want both! High resource utilization and low recovery delay
Dynamic re-routing in mesh networks
Dynamic re-routing in Mesh network
A disjoint back-up route is pre-calculated, but not
allocated
Routing table has entries for backup route
Hardware signaling helps to set up backup path rapidly
May be blocked
F
E
A
D
B
C
Analysis and simulation (ongoing work)
A 13-node, 22-link
sample network,
average number of
hops on primary path
is 2.4, on secondary
path is 3.5.
Analysis of dynamic
provisioning (M/D/1/∞)
Tsig E (h)(Tsig
Tsig
2(1 Tsig )
Tsig )
• Simulation of dynamic re-routing
• When link/node fails, all affected paths must be re-routed
simultaneously. We use OPNET to model the dynamic rerouting behavior.