2. 2 T1X1.5/2002-046
CONTENTS
• What is GFP
• Frame format of GFP
• Modes of GFP
• Types of GFP
• Packet routing
• Resilient Packet rings
• Extending LAN/SAN over WAN
• SAN Transport
• Transparent GFP Mapping
• Frame vs. Transparent GFP
• Conclusion
3. 3 T1X1.5/2002-046
What is GFP?
• GFP is an emerging new standard for Data
Encapsulation
• Accepts any client, encapsulate in simple frame,
transport over network
• Uses length/HEC frame of variable length packets
• Allows multiple data streams to be transported
over single path
– Packet aggregation for router applications
– Common encapsulation of different client data
types (e.g. Ethernet, HDLC)
4. 4 T1X1.5/2002-046
• ITU standard, G.7041 describes a Generic
Framing Procedure (GFP) which may be
used for efficiently mapping client signals
into and transporting them over
SONET/SDH or G.709 links. This
presentation provides an overview of
network applications which have driven the
development of the GFP standard.
Applications are related to some of the
features included in G.7041.
5. 5 T1X1.5/2002-046
• Transparent Mapping supports LAN/SAN
extension over WAN
• Extension headers support various
network topologies
–Null Extension Header for channelized
Point-to-Point network
–Linear Extension Header for Port
Aggregation over Point-to-Point network
–Ring Header for Resilient Packet Ring
applications.
6. 6 T1X1.5/2002-046
FRAME FORMAT
A GFP frame consists of:
• A core header
• A payload header
• An optional extension header
• A GFP payload
• An optional payload frame check
sequence
8. 8 T1X1.5/2002-046
MODES OF GFP
•There are two modes of GFP:-
• Generic Framing Procedure
- Framed (GFP-F)
• Generic Framing Procedure
- Transparent (GFP-T)
9. 9 T1X1.5/2002-046
• GFP-F maps each client frame into a
single GFP frame. GFP-F is used where
the client signal is framed or packetized
by the client protocol.
• GFP-T on the other hand, allows
mapping of multiple 8B/10B block-coded
client data streams into an
efficient 64B/65B block code for
transport within a GFP frame.
10. 10 T1X1.5/2002-046
TYPES OF GFP
There are two types of GFP frames:-
•A GFP client frame
•A GFP control frame
11. 11 T1X1.5/2002-046
• A GFP Client frame can be further
classified as either a client data frame or
a client management frame. The former
is used to transport client data, while the
latter is used to transport point-to-point
management information like loss of
signal, etc
12. 12 T1X1.5/2002-046
• The GFP Control frame currently
consists only of a core header field with
no payload area. This frame is used to
compensate for the gaps between the
client signal where the transport medium
has a higher capacity than the client
signal, and is better known as an idle
frame.
13. 13 T1X1.5/2002-046
Application: Packet Routing through Big Fat Pipes
Packet
Switch
N x GbE
SONET
SDH
Mapper
SONET
SDH
Mapper
SPI-4
SPI-3
Router-based
WAN
OC-48
STM-16
OC-192
STM-64
• Packet Switch encodes/decodes 8B/10B and routes packets to appropriate SPI-n
• SONET/SDH Mapper encapsulates packets using PPP over GFP and maps them
into concatenated payload (STS-48c/VC-4-16c or STS-192c/VC-4-64c)
• All packet switching in WAN is handled by Layer 2 routing
• Single traffic type aggregated in edge switch & routers into big-fat-pipes going to
desired hop in routing table
• Control info from 8B/10B encoding is not preserved
• Relies on PPP for Link Configuration
Edge Switch
15. 15 T1X1.5/2002-046
Application: Resilient Packet Rings
GbE
MAC
OC-m
STM-n
Packet
Ring
HDLC
Proc.
SONET
SDH
Mapper
Framer
• Resilient Packet Ring (RPR), also known as IEEE 802.17, is a protocol
standard designed for the optimized transport of data traffic over optical fiber
ring networks.
• Multiplex packet streams into single STS-Nc / VC-4-Xc
• Each packet is encapsulated into GFP Frame
• Payload Type ID in payload header supports multi-service applications.
Network
Process.
&
Switch
SPI-nSPI-n
8B/10B
Client
Packet
Stream
Ring Node Ring
Node
Ring
Node
Ring
Node
Packet Add/Drop
16. 16 T1X1.5/2002-046
GFP Frame: RPR Using GFP Ring Header
Core Header
Payload
Area
Length MSB
Length LSB
cHEC MSB
Packet
Payload
Payload Header
Payload Type MSB
Payload Type LSB
tHEC MSB
tHEC LSB
DestPort SrcPort
Spare
Spare DE CoS
TTL
Dest MAC[47:40]
Dest MAC[39:32]
Dest MAC[31:24]
Dest MAC[23:16]
Dest MAC[15:8]
Dest MAC[7:0]
Src MAC[47:40]
Src MAC[39:32]
Src MAC[31:24]
Src MAC[23:16]
Src MAC[15:8]
Src MAC[7:0]
eHEC MSB
eHEC LSB
Ring
Extension
Header
cHEC LSB
FCS (optional)
FCS[31:24]
FCS[23:16]
FCS[15:8]
FCS[7:0]
NOTE: GFP Ring Header removed to Living
List; 802.17 RPR proposes to include ring
header as part of GFP payload).
17. 17 T1X1.5/2002-046
Application: Extending LAN / SAN over WAN
GbE
FC
8B/10B
Clients
STS-m
STM-n
8B/10B
Client
STS-m
STM-n
SONET / SDH
Network
GbE
FC
GbE
FC
GbE
FC
LAN /
SAN
8B/10B
Client GbE
FC
SONET
SDH
Mapper
Framer
SONET
SDH
Mapper
Framer
SONET
SDH
Mapper
Framer
• If you want to preserve individual 8B/10B block-coded channels, but cannot fit
two1.25 Gb/s GbE channels into a single OC-48 / STM-16
• Transport of single 1.25 Gb/s stream over OC-48 / STM-16 is excessively wasteful.
• Need to preserve control info (e.g. link configuration) for LAN extension, so we
cannot just send data packets.
• Cannot just interleave two streams into single path and still expect SONET/SDH to
deliver to different destinations.
18. 18 T1X1.5/2002-046
SAN Transport through Right-Sized Pipes using GFP
N x
Fibre Chan,
GbE,
FICON,
ESCON
SONET
SDH
Mapper
with VC
SONET/SDH
Switched
WAN
OC-48/STM-16 or
OC-192/STM-64
• Transparent Encapsulation / Decapsulation preserves Control Info
• Virtually-concatenated (VC) paths sized to fit individual client signals
• Client signals preserved intact through the network
• Signals routed by switching VC paths (STS-1/VC-3 or STS-3c/VC-4 switching)
• Mix of protocols may be carried, each in its own VC path
• Virtual Concatenation (VC) is essential to compete against SAN over dark fiber
SAN - WAN PHY
8B/10B
Codec
8B/10B
Codec
Transparent
Encapsulate
/ Extract
Transparent
Encapsulate
/ Extract
20. 20 T1X1.5/2002-046
Frame-Mapped GFP vs. Transparent GFP
Frame-Mapped GFP Transparent-Mapped GFP
Variable Length GFP Frames Fixed Length GFP Frames
1-to-1 mapping of Data Packets to GFP
Frames
N-to-1 mapping of client “characters” to GFP
Frames
Point-to-Point, Packet Aggregation, or
Resilient Packet Ring Network Topology
Primarily Point-to-Point Topology using Virtual
Concatenation
Requires “MAC” to terminate client signal and
pass only data packets.
Only 8B/10B PHY layer terminated; “MAC”
not required to terminate higher layer protocol.
Data only passed in 8B format. Data and control compressed using 64B/65B
re-coding.
Channel-associated control possible using
GFP Control Frames.
Channel-associated control possible using
GFP Control Frames.
Unclear if client Loss-of-Sync, or code
violations should be communicated to far-end.
Transparent mapping defines mechanisms for
communicating LOS, Loss-of-Sync, code
violations to far end.
Doesn’t define client egress action due to
SONET/SDH signal failure.
Defines client egress action due to
SONET/SDH signal failure.
21. 21 T1X1.5/2002-046
GFP Conclusion
• Various GFP Applications have been described and illustrated
– Packet routing
– Port aggregation over SONET/SDH using Linear Extension Headers
– Resilient Packet Ring applications using Ring Extension Headers
– Transparent Transport of 8B/10B clients
• Basic GFP Frame Structure has been described and shown
– Length frame delineation, similar to ATM cell delineation.
– Payload Headers ID encapsulated payload & encapsulation options
• Presence or absence of optional FCS
• Presence and type or absence of extension header
• Payload type allows for mixing data types in a single SONET/SDH
– Extension headers support various network topologies
• Null Extension Header for channelized Point-to-Point network
• Linear Extension Header for Port Aggregation over Point-to-Point
network
• Ring Header for Resilient Packet Ring applications
22. 22 T1X1.5/2002-046
• LAN/SAN extension over WAN using Transparent
Mapping described and shown
– 64B/65B re-coding preserves data & control for
“transparent” transport
– Superblocks provide error detection / correction
over relatively small blocks
– Supports efficient transport of full-rate 8B/10B
clients over smallest paths.