Professional Documents
Culture Documents
Nortel Networks is an industry leader and innovator focus ed on transforming how the world communicates and exchanges
information. The company is supplying its service provider and enterprise customers with communications technology and
infrastructure to enable value-added IP data, voice and multimedia services s panning Wireline, Wireless Networks,
Enterprise Networks, and Optical Networks. As a global company, Nortel Networks does business in more than 150
countries. More information about Nortel Networks can be found on the Web at:
www.nortelnetworks.com
August 2003
White Paper
Introduction
Given the current economic climate of the embattled telecom industry, service providers are
diligently seeking out new methods to significantly reduce their operational costs, while
simultaneously offering new revenue generating data services. These emerging data services
must fully leverage, and not strand, the huge installed base of SONET/SDH networking
equipment to facilitate a cost-effective migration path. Although numerous technologies to carry
data services exist, the use of certain interrelated technologies is gaining in popularity within the
metro networking market space. In particular, this paper discusses the suite of technologies listed
below as one possible method of meeting the often contradicting goals of introducing new
revenue generating services at a reduced cost.
These technologies complement one another and together facilitate such data services as GbE
(Gigabit Ethernet), FC (Fibre Channel), FICON (Fiber Connection), ESCON (Enterprise System
Connection), IP (Internet Protocol), and PPP (Point-Point Protocol) in the MAN (Metro Area
Network) at a reduced complexity and lower operational cost when compared to existing methods
of transporting these increasingly popular data services over SONET/SDH.
2 of 11
White Paper
16-Bit Payload
Length Indicator
Core Header
cHEC (CRC-16)
Payload Headers
(4 to 64 Bytes)
Type
(4 Bytes
Extension
(0 60 Bytes)
Payload Area
Optional Payload
FCS (CRC-32)
PLI (2 Bytes)
cHEC (2 Bytes)
FCS (4 Bytes)
3 of 11
White Paper
the available virtual container bandwidth within SONET/SDH and OTN. Since at least eight
8B/10B characters must be received, decoded, and then encoded once again into a 64B/65B
super block, this processing sequence becomes the minimal latency of GFP-T, which is far less
than that of GFP-F. Framed GFP waits for an entire Ethernet frame to be received, which is
variable in size from 64 to 1518 bytes, resulting in increased latency, which is unsuitable for FCbased networks. As its name implies, GFP-T transparently transports 8B/10B control characters
and data characters in an agnostic manner which is another notable advantage of GFP-T over
GFP-F for 8B/10B encoded data services.
Alternative Data Services Mapping Techniques
There are competing alternatives to GFP mapping for carrying data services over SONET/SDH
based networks such as PoS (Packet over SONET/SDH) [11] and X.86 [12] over SDH. However,
when compared to GFP, these alternative mapping schemes lack the flexibility and simplicity of
GFP and are thus not as desirable a choice for wide-scale deployment. Nevertheless, they are
functional, albeit less efficient, alternatives available today.
PoS (Packet over SONET/SDH) Data Services Mapping
PoS is currently the most common method of mapping Ethernet/IP data services over
SONET/SDH networks via router interfaces that support HDLC (High Level Data Link Control)
[13], which are widely available. However, unlike PoS, GFP does not truncate the L2 MAC
information (e.g. destination/source addresses) of an Ethernet frame thus enabling transparent
connections of L2 edge devices such as Ethernet switches. PoS terminates this L2 MAC
information, which prohibits numerous desirable L2 features such as multicasting, traffic
prioritization (802.1p), Ethernet protection switching, and VLAN (Virtual LAN) filtering (802.1q).
PoS will strip this valuable L2 MAC information, and then subsequently remap it into PPP over
HDLC which creates, in effect, a dumb point-to-point pipe with some significant disadvantages
that are discussed below.
HDLC relies on specific bit patterns (characters) that are used for frame delineation and control
information. Thus, these specific bit patterns, if found within the data payload, would inadvertently
mimic reserved characters and subsequently interfere with the proper operation of HDLC. To
overcome this issue, HDLC injects additional control escape characters (0x7d) adjacent to the
payload bit patterns that mimic reserved characters. While addressing one issue, however, the
addition of control characters introduces another: increased bandwidth is required to carry the
original client information. This is commonly (and rather unflatteringly) referred to as bandwidth
inflation which is non-deterministic. In the worst case, if the client signal is composed solely of
data characters that inadvertently mimic reserved control characters, HDLC used in PoS would
inject escape characters on a one-to-one basis, thereby doubling the bandwidth required to
carrying the original client signal an undesirable (yet quite unlikely) scenario.
To avoid bandwidth inflation that is inherent to HDLC and thus PoS, GFP uses its cHEC (core
Header Error Control) bytes for frame delineation instead. This use of cHEC bytes is actually an
adaptation of the proven ATM cell delineation algorithm and is illustrated in Figure 3. The state
machine illustrates how GFP starts off in the Hunt state and scans bit by bit until a calculated
cHEC matches a subsequent sequence of received bits and results in the delineation of a GFP
frame whereby the Pre-Sync state is reached. Since one GFP frame is successfully delineated,
the next GFP frame starting point is known. If a second GFP frame has a valid cHEC and
subsequent sequence of bits, the Sync state is entered. The Sync state allows for cHEC error
correction but if the received error is not correctable, the Hunt state is reentered and the process
repeats itself. As no special characters are required for GFP to delineate received frames, the
adverse bandwidth inflation issue that is apparent in PoS/HDLC is thus eliminated.
4 of 11
White Paper
Pre-Sync
State
cHEC Match
No 2 nd cHEC Match
Hunt
State
Sync
State
No cHEC Match
Standardization this will accelerate global acceptance for vendor interoperability and lower
component costs
Versatility GFP enables the transport of many popular data service protocols at L1 and L2.
o L1 GbE, FC, FICON, ESCON, Infiniband (future), and DVB ASI (future).
o L2 Ethernet (e.g. services such as IP/PPP/MPLS encapsulation) and HDLC.
Quality of Service (QoS) strict (GFP-T) or loose (GFP-F) QoS is possible depending upon
the application.
Scalability GFP currently supports data services from 10Mb/s (Ethernet) to 10Gb/s (OC192/STM -64).
Acceptance GFP is endorsed by the IEEE-802.17 RPR (Resilient Packet Ring) Working
group and the IETF.
Simplicity GFP is a simpler mapping technique than ATM and HDLC resulting in less costly
network designs.
5 of 11
White Paper
Header Error Control (HEC) it is based upon a scheme similar to the proven ATM algorithm
and supports fixed (Transparent) and variable (Framed) client frame sizes for added
flexibility. The scheme is not process intensive and is thus conducive to hardware integration
for high speed operation at a reduced cost.
Although GFP has many inherent benefits, it is not sufficient for enabling the end-to-end transport
of data services that are increasingly migrating into the MAN. Other technologies are required to
size the needed SONET/SDH payload carrying capacity, and to dynamically and hitlessly resize
bandwidth. These requirements are addressed by Virtual Concatenation and the Link Capacity
Adjustment Scheme, respectively. Another layer of optimization is addressed by Resilient Packet
Ring with Coarse Wavelength Division Multiplexing determining the overall bandwidth capacity.
These technologies and their interrelation are discussed in subsequent sections of this paper.
SONET/SDH
Contiguous Concatenation
STS-3c (67%)
STS-48c (42%)
STS-12c (27%)
STS-48c (35%)
SONET/SDH
Virtual Concatenation
STS-1-2v (100%)
STS-1-21v (95%)
STS-1-4v (82%)
STS-3c-6v (95%)
Table 1 Comparison of Contiguous vs. Virtual Concatenation (VCAT) Data Services Mapping
Only the end points (network ingress and egress points) of a VCAT must be aware of the
relationship between the constituent VCG members. Intermediary nodes need not be aware of
the relationship between VCG members which greatly facilitates the cost-effective introduction of
VCAT into existing SONET/SDH networks. Only the ingress/egress network nodes need be
VCAT enabled. Since VCAT members do not have to follow the same multiplex section,
propagation delays between VCG members can result, especially over long distances. However,
VCAT is able to identify these inevitable propagation delays and perform the necessary
realignment process.
6 of 11
White Paper
All new technologies have associated shortcomings and VCAT is no exception. For instance, if
the path of one of the constituent VCG members fails, the entire VCG fails. VCAT also does not
solve the point-point nature of SONET/SDH which is not conducive to L2 protocols such as
Ethernet. Although one can provision multiple SONET/SDH connections in the form of a mesh
network, this is costly, time-consuming, and results in significant operational and capital
expenditures. Although VCAT improves the rigid nature of a SONET/SDH circuit, it still results in
a non-optimal solution since carriers must provision circuits based on the peak rates of
customers bursty data services regardless of whether the connections are contiguous or noncontiguous. The average utilization rate may actually be much lower, resulting in stranded
bandwidth a rather undesirable suboptimal solution.
LCAS complements VCAT to overcome the loss of an entire VCG when one of the group member
paths fails by enabling a dynamic hitless addition/deletion of individual group members. Since
VCAT is essentially transparent to L2 data services, RPR can be used to overcome VCAT the
deficiencies due to its point-to-point nature. RPR enables a logical mesh L2 network topology
within a physically constructed SONET/SDH ring architecture thus enabling the leveraging of
already deployed SONET/SDH infrastructure.
7 of 11
White Paper
deployed OSS (Operational Support Systems) products, and initiatives to incorporate these
features are underway. LCAS, however, does not actually participate in the provisioning process
of such new VCAT circuits. The NMS/EMS initiated creation and resizing of VCAT connections
require automated signaling to the VCAT/LCAS enabled network nodes through a generalized
signaling protocol such as MPLS (Multi-Protocol Label Switching), discussed later in this paper.
Other standardized signaling and control protocols may also be used.
8 of 11
White Paper
standardized signaling/control protocols to initiate active network reconfigurations based on
changing service demands. One such protocol is MPLS. Since RPR is actually a L2 technology, it
can also transport MPLS signaling and control information which would be oblivious to the
physical nature of the RPR with each node actually appearing to be connected in a fully meshed
configuration. RPR can use GFP or HDLC to encapsulate RPR packets over a SONET/SDH
based network infrastructure, although GFP is the preferred method since it does not suffer from
the detrimental HDLC limitations (e.g. bandwidth inflation) already discussed.
However, there are certain caveats related to RPR worth mentioning. The most important issue is
that RPR has not yet been fully standardized meaning the solutions deployed today may not be
compliant to the standards that will eventually be finalized. There is also the additional overhead
of RPR which is over and above the Ethernet overhead being encapsulated. Fortunately, this
added overhead pales in comparison to the achieved RPR benefits.
9 of 11
White Paper
aforementioned technologies, but initiatives are underway to accelerate their adoption due to the
obvious benefits brought forth.
Emerging Applications
There are numerous applications emerging in the metropolitan optical networking industry that
involve high-speed data services (i.e. Ethernet and Fibre Channel) requiring long reach at a cost
that is economically viable. One such application is the SAN that essentially geographically
separates the application network from the data storage network. This allows for the extension
and subsequent centralization of the data storage facilities to allow numerous physically
separated applications (e.g. transaction processing) to share the same centralized data storage
facility in real-time. The current SAN technology of choice is leaning towards Fibre Channel with
native optical reaches of 500m over multimode fiber and ~10km over single-mode optical fiber,
which although admirable, still greatly limits the addressable application space. New technologies
are required to further extend the reach of Fibre Channel reach which can be effectively
addressed using a judicious mix of GFP, VCAT, LCAS, and SONET/SDH technologies.
One specific application involving a geographic separation of computing/storage facilities is a
direct result of the recent tragic events that occurred in September 2001. It brought to the
forefront the issue of security and disaster recovery plans to ensure business continuity in the
event of a catastrophic event that could cripple and even bankrupt a business and/or industry. As
a result, the major American financial regulatory agencies (Federal Reserve and Securities &
Exchange Commission) discussed how to ensure business continuity in the event of a disaster,
and produced guidelines detailed in the Draft Interagency White Paper on Sound Practices to
Strengthen the Resilience of the U.S. Financial System [14]. One of the key suggestions is to
physically separate the duplication (standby) of critical business facilities and operations by 200300 miles, which mandates data networking technologies that today are not capable of
addressing in an economical manner. The current lack of cost-effective data services mapping
into long haul core networks is one area that has many in the industry feeling uncomfortable with
the proposed guidelines. However, the combination of the aforementioned technologies, in
particular, mapping Fiber Channel and Ethernet services onto the reach capabilities of
SONET/SDH, can help to alleviate these significant concerns.
Going Forward
Standardization of any technology quickly leads to commercially available chip sets, increased
market adoption rates, and subsequent lower costs which in turn spur on market adoption.
Ethernet is just one such example of a technology that due to its standardized nature and rapid
market acceptance, has resulted in an extremely cost-effective and ubiquitous technology. The
technologies discussed in this paper are for the most part standardized, or nearly standardized,
and will lead to the next generation of SONET/SDH networks and services. Additional data
services will continue to migrate deeper into existing SONET/SDH MANs (and WANs) thus
creating new revenue generating data services while simultaneously leveraging this huge
installed network base. The demarcation points between LANs and MANs are increasingly
blurring creating the urgent business need to marry the two technologies of choice Ethernet and
SONET/SDH. The judicious introduction of the aforementioned technologies is one available
option to ensure that these ever-blurring demarcation points are ultimately removed.
10 of 11
White Paper
Acknowledgement
The author would like to thank Thomas Barnwell for his contribution to this conference paper
References
[1]
[2]
11 of 11