Network Working Group Luc Ceuppens (Chromisys)
Internet Draft Dan Blumenthal (Chromisys)
Expiration Date: September 2000 John Drake (Chromisys)
Jacek Chrostowski (Cisco Systems)
W.L. Edwards (iLambda Networks)
Performance Monitoring in Photonic Networks
in support of MPL(ambda)S
draft-ceuppens-mpls-optical-00.txt
Status of this Memo
This document is an Internet-Draft and is in full conformance with
all provisions of Section 10 of RFC2026.
Internet-Drafts are working documents of the Internet Engineering
Task Force (IETF), its areas, and its working groups. Note that
other groups may also distribute working documents as Internet-
Drafts. Internet-Drafts are draft documents valid for a maximum of
six months and may be updated, replaced, or obsoleted by other
documents at any time. It is inappropriate to use Internet- Drafts
as reference material or to cite them other than as "work in
progress."
The list of current Internet-Drafts can be accessed at
http://www.ietf.org/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
1. Abstract
Realizing the important role that photonic switches can play in
data-centric networks, work has been going on within the IETF to
combine the control plane of MPLS (more specifically traffic
engineering) with the point-and-click provisioning capabilities of
photonic switches [1]. This document outlines a proposal to expand
this initiative to include DWDM, OADM and ATM systems. It also
proposes to expand the work beyond simple establishment of optical
paths and include optical performance monitoring and management. The
combined path routing and performance information that will be
carried and shared between these network elements will allow the
elements or element management system (EMS) to adequately assess the
"health" of an optical path (which can be a wavelength or fiber
strand). The routers and/or ATM switches at the edges of the
photonic network will then use this information to dynamically
manage the millions of wavelengths available in the photonic layer.
Ceuppens/Blumenthal et al. [Page 1]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
2. Introduction
Over the last two years, DWDM has proven to be a cost-effective
means of increasing the bandwidth of installed fiber plant. While
the technology originally only served to increase the size of the
fiber spans, it is quickly becoming the foundation for networks that
will offer customers a new class of high-bandwidth and broadband
capabilities.
Sales of DWDM systems will reach $6 billion in North America by the
end of 2000. This roughly translates into tens of thousands of
wavelengths deployed within optical networks, either as point-to-
point connections or in ring topologies. In addition, several
millions of wavelengths are projected to be deployed in enterprise,
metropolitan, regional, and long haul networks by 2007 in the United
States alone.
These wavelengths will require routing, add/drop, and protection
functions, which can only be achieved through the implementation of
network-wide management and monitoring capabilities. Current-
generation DWDM networks are monitored, managed and protected within
the digital domain, using SONET and its associated support systems.
However, to leverage the full potential of wavelength-based
networking, the provisioning, switching, management and monitoring
functions have to move from the digital to the optical domain.
Efficiently managing (i.e., adding, dropping, routing, protecting,
and restoring) the growing number of traffic-bearing wavelengths can
only be achieved through a new breed of optical networking element.
This network element is the photonic switch*.
Photonic switching is the next logical step in a long history of
switching technology that started with manual "plug board"
operators, evolved to mechanical crossbar and finally digital
switching. Photonic switching will enable transparent photonic
networks. Photonic networks will greatly simplify the architecture
of both the network and the network nodes by establishing end-to-end
optical paths across the network. An end-to-end photonic path
behaves as a transparent$ "clear channel", so that there is
virtually nothing in the path to limit the throughput of the fibers.
* Photonic switches are often referred to as optical cross-connects
(OXC). However, today's OXCs are based on electrical rather than
photonic switching fabrics, and therefore do not demonstrate the
optical transparency required to grow photonic networks in the
future. We use the term photonic switch to distinguish these classes.
$ Transparency implies that signals with any type of modulation schemes
(analog or digital), any bit rate, and any type of format can be
superimposed and transmitted without interfering with one another,
and without their information being modified within the network.
Opaque networks do not have this property.
Ceuppens/Blumenthal et al. [Page 2]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
A transparent channel essentially behaves like an ideal
communications with almost no noise and very large bandwidth.
Secondly, as the nodes in a photonic network have essentially no
data processing to do, they can be made extremely simple and hence
very cheap. Finally, optical node simplicity also means simplicity
of control and management.
Without any doubt, the next revolution in the telecommunications
industry will occur within the optical domain. Now that the basic
components are available to build photonic networks, the most
important innovations will come from adding intelligence that
enables the interworking of all the network elements (Routers, ATM
switches, DWDM transmission systems and photonic switches). This new
photonic internetwork will make it possible to provision high
bandwidth in seconds, turning the new optical technology into a
revenue spinner for the service provider rather than just a way of
saving money.
However, the intelligent open photonic network can only be built if
the currently vertically layered network migrates to a horizontal
model where all network elements work as peers to dynamically
establish optical paths through the network. The IETF has already
addressed the interworking of routers and optical switches through
the MPL(amda)S initiative [1]. We propose to expand this initiative
to include DWDM systems and ATM systems. We also propose to expand
the work beyond simple establishment of optical paths and include
optical performance monitoring and management. The combined
information that will be carried and shared between these network
elements will allow the elements or element management system (EMS)
to adequately assess the "health" of an optical path (which can be a
wavelength or fiber strand). The routers and/or ATM switches at the
edges of the photonic network will then use this information to
dynamically manage the millions of wavelengths available in the
photonic layer.
As a summary, the following functions need to be covered
1. Dynamic Bandwidth Provisioning
2. Optical Performance Monitoring
3. Signaling for 1 and 2.
The remainder of this document discusses these functions into more
detail.
Ceuppens/Blumenthal et al. [Page 3]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
3. Dynamic bandwidth provisioning
As indicated above, the photonic network uses photonic switches and
optical transmission equipment to provide point-to-point connections
to attached internetworking devices. These connections will
typically take the shape of dedicated wavelengths, but can also be
SONET leased line services or gigabit Ethernet connections. While
the photonic network will typically provide these bandwidth services
to IP routers, the model should be extended to include ATM switches.
While the idea of bandwidth-on-demand is certainly not new, existing
networks do not support instantaneous service provisioning. Current
provisioning of bandwidth is painstakingly static. Activation of
large pipes of bandwidth takes anything from weeks to months.
The imminent introduction of photonic switches in the transport
networks opens new perspectives. Combining the bandwidth
provisioning capabilities of photonic switches with the traffic
engineering capabilities of MPLS [2], will allow routers and ATM
switches to request bandwidth where and when they need it.
To make this work, however, requires more than simply advertising
the availability of routes by the photonic switches to the routers
and/or switches. They will also need to provide information about
the characteristics and performance of the paths. Adequately
assessing the status and health of an optical path through the
photonic network requires a detailed cooperation between the
photonic switches and the transmission systems providing the basic
transport capabilities in the long-haul network.
4. Performance Monitoring
Service providers to date have limited the role of DWDM in the
network to creating "virtual fiber", i.e., the straightforward
increase in capacity of the fiber plant, even if this meant a
dramatic increase in complexity since each virtual fiber required
the deployment of its own SONET equipment. The reason behind this
restricted role is the worry about network management, alarm
monitoring and protection capabilities of DWDM systems and the
photonic layer in general.
Current performance monitoring in optical networks requires
termination of a channel (wavelength) at an OEO (optical-electrical-
optical conversion) point to detect bits related to BER of the
payload or frame (e.g., SONET LTE monitoring). For example, one form
of error checking can be carried out at the SONET level by
monitoring the B1 and J0 overhead bytes of the SONET stream.
However, while these bits indicate if errors have been received,
they do not supply channel-performance data. This makes it very
difficult to assess the actual cause of the degraded performance.
Ceuppens/Blumenthal et al. [Page 4]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
The premise of photonic networks requires the availability of tools
to measure and control the smallest granular component of such
networks --the wavelength channel. These functions include the
monitoring of amplifiers and switches at add/drop sites, the
deployment and commissioning of DWDM routes, as well as the
restoration and protection of networks. This must be accomplished
with speed and accuracy over an extended period of time.
Fast and accurate determination of the various performance measures
of a wavelength channel implies that measurements have to be done
while leaving it in optical format. In the remainder of this
document we will refer to this as "optical performance monitoring"
(OPM). One possible way of achieving this is by tapping a portion of
the optical power from the main channel using a low loss tap of
about 1%. In this scenario, the most basic form of OPM will utilize
a power-averaging receiver to detect loss of signal (LOS) at the
optical power tap point. Existing DWDM systems use OTDRs (Optical
time-domain reflectometers) to measure the parameters of the optical
links.
As photonic networks mature, it will be desirable to generate a more
detailed picture of the channel "health" in a manner that can be
communicated to the EMS and other network control entities, as well
as between other network elements. By monitoring various OPM
parameters, one can attempt to estimate the BER, detect gradual or
sudden performance degradation, and report these to local or global
NMS entities, and to attached internetworking devices. Also, fiber
spans are typically characterized, or calibrated, during the
provisioning process on DWDM systems, as fiber manufacturer, fiber
type etc. all have a bearing on how the various DWDM spectrums
should be populated. It would be useful to have the calibrated data
for each fiber span available as part of the overall information on
the photonic layer. All the available information can then be
correlated across the network to make decisions on fault isolation
and take appropriate actions such as rerouting the connection or
adaptively downgrading or upgrading the bit-rate of a channel.
When deploying an optical network it is common practice to document
the baseline for all operating parameters, such as signal power,
bit-error rate, OSNR, etc., prior to network turn-on. During normal
operation, network elements equipped with OPM capabilities can
report any degradation events of the optical channel to the network
operations center (NOC) and to the other network elements. The
element management system (EMS) can document the degradation of the
photonic layer in time by saving optical performance monitoring data
in an archival database. As channels are added, removed or rerouted,
the NOC can continuously monitor and analyze the status as channels
are dynamically managed.
With the advent of an open photonic network, we can envision a trend
of leasing channels or wavelengths that span multiple networks. This
will require optical interconnects between various networks.
Ceuppens/Blumenthal et al. [Page 5]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
Invariably, as channels are handed off between carriers, problems
can occur which require monitoring to resolve conflicts. Most of
these issues occur at network boundaries. In addition, if service
providers offer various levels of QoS, both networks will have a way
of negotiating the end-to-end QoS of the channels and both service
providers will need a way to ensure that the other party lives up to
the service contract. Here again, independent monitoring is needed
to ensure quality and continuity of service.
The issue of effective OPM sensitivity will impact how pervasive
each technique is used in a network due to cost and complexity.
Certain techniques may require an optical amplifier at the tap point
resulting in OPM module sensitivity equivalent to that of the final
path termination point. Other issues that need to be addressed
include definition of OPM at the section, line and path levels.
Since monitoring can be in principal performed at any point within
the network, traditional use of LTE points does not carry over.
Another problem related to transparency lies in determining the
threshold values for the various parameters at which alarms must be
declared. Very often these values depend on the bit rate on the
channel and should ideally be set depending on the bit rate.
However, in a truly transparent network, one may have to set alarms
to correspond to the highest possible bit rate that can be present
on a channel. In addition, since a signal is not terminated at an
intermediate node, if a wavelength fails, all nodes along the path
downstream of the failed wavelength could trigger an alarm. This can
lead to a large number of alarms for a single failure, and makes it
somewhat more complicated to determine the cause of the alarm (alarm
correlation).
We see as potential candidates, the following OPM functions:
1. Dispersion (chromatic and polarization mode):
The distortion or spreading of bits due to variations in
propagation velocity of different wavelengths and polarization
modes in the fiber and other network elements.
2. Optical signal-to-noise ratio (OSNR):
The ratio of optical power in a primary data channel to the power
in optical background noise accumulated during transmission and
switching. This ratio is usually specified within some optical
bandwidth of a receiver filter. The OSNR of a channel at the
destination receiver will set the limit of the final detected SNR.
3. Bit-rate
The data rate of the channel in a transparent system will be
necessary to make decisions on other performance metrics.
4. Q-Factor
A measure of the signal-to-noise ratio (SNR) assuming Gaussian
noise statistics.
Ceuppens/Blumenthal et al. [Page 6]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
5. Wavelength registration
The determination of which wavelengths are present on a given
fiber.
6. Wavelength selective component drift
The drift of a laser, filter, mux or other wavelength selective
component relative to the ITU grid.
7. Optical cross talk
Two types of cross talk are of interest, in-band and out-of-band.
In-band cross talk is seen as at the same wavelength as the
primary channel and appears as cross talk in the electronic
domain. Out-of-band cross talk appears as a different wavelength
in the presence of the primary wavelength and appears as cross
talk in the optical domain.
8. Optical power transients
Changes in the optical powers that are not due to normal bit
transitions. May be due to optical amplifier gain transients or
other transient non-linearity in the system.
9. Bit-error-rate
In a SONET environment BER can be directly measured on the channel
using means to look at bits within the data stream. However, in a
purely photonic network there will typically not be access to the
data streams carried over the channel. However, by interpreting
the other optical parameters, the system should be able to
estimate the BER with relatively good accuracy, as well as
guarantee bit error rate performance to the users of the channel.
10.Jitter
Random fluctuations in the location of rising and falling edges of
bits relative to a local or recovered clock reference. As line
speeds continue to increase, jitter will become a critical
performance parameter.
11.Insertion loss
Indicates the input to output loss of a network element. When
examining excessive power loss along the path of a channel the
ability to measure insertion loss of individual network elements
is very useful, specifically when compared against an archival
database.
12.Optical power level
In addition to verifying the service level provided by the network
to the user, performance monitoring is also necessary to ensure that
the users of the network comply with the requirements that were
negotiated between them and the network operator. For example, one
function may be to monitor the wavelength and power levels of
Ceuppens/Blumenthal et al. [Page 7]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
signals being input to the network to ensure that they meet the
requirements imposed by the network.
To make any Performance Measurement metrics meaningful, major effort
should be on conducting serious testing to draw correlation between
the proposed Optical measurement metrics with the quality of the
signals (electrical).
5. Signaling
The vast majority of existing communications networks uses framing
and data formatting overhead as the means to communicate between
network elements and management systems.
It is clear however, that truly transparent and open photonic
networks can only be built with transparent signaling support.
Arguments in favor of transparency include, but are certainly not
limited to:
- Framing and formatting makes the network opaque and as such
inhibits the creation of bit rate and protocol transparent
networks. As overhead information is processed in the digital
domain, it requires an optical-to-electrical and electrical-to-
optical conversion at every point in the network where traffic is
inserted or dropped and at each point where management and
monitoring is required. This imposes severe limitations and is
probably the single biggest inhibitor of growth in current optical
networks. That is why "digital wrappers" are not a viable
solution. In fact, an all-optical network using digital wrappers
is a contradiction in terms!
- Attached internetworking equipment and customer equipment may not
support the framing overheads.
- In today's optical network (I.e., SONET) the service and
infrastructure layer are inseparable. As a result, "optical-
network-ignorant" protocols such as (ten) gigabit Ethernet, fiber
channel or ESCON cannot be transported without being translated to
the infrastructure layer. Hence the need for adaptations such as
"gigabit Ethernet over SONET", "packet over SONET" etc.
In contrast, separate control plane techniques# supports flexible
control and management of multi-vendor networks and will pave the
way for truly open and transparent photonic networks.
# There may be instances where some "embedded" wavelength routing
information is required. One such instance is in existing networks
where DWDM junctions are "hard-wired" and the end-to-end path may
consist of different wavelengths.
Ceuppens/Blumenthal et al. [Page 8]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
As already mentioned, the approach proposed by the MPLS/TE [1] task
force of IETF propose to combine recent advances in MPLS traffic
engineering control plane with emerging photonic switching
technology to provide a framework for real-time provisioning of
optical channels and allow the use of uniform semantics for network
management operations control in hybrid networks consisting of
photonic switches and label switching routers (LSRs). While the
proposed approach is particularly advantageous for data-centric
optical internetworking systems, it can easily be expanded to
include basic transmission services. Similarly, it can be expanded
beyond simple bandwidth provisioning to include optical performance
monitoring.
It is worth mentioning that while the signaling is used to
communicate all monitoring results, the monitoring itself is done on
the actual data channel, or some range of bandwidth around the
channel. Therefore, all network elements must be guaranteed to pass
this bandwidth in order for monitoring to happen at any point in the
network.
Several signaling flows have to be supported:
- between the internetworking equipment and the photonic cross-
connect
- between the photonic cross-connect and the DWDM transmission
systems
- between the DWDM systems and optical amplifiers
- between the DWDM systems and optical add/drop multiplexers
- between the internetworking devices and optical add/drop
multiplexers or DWDM transmission systems (if this connection does
not run through a PXC)
We propose that the connection signaling be limited to exchanges
between the internetworking device and the transmission network
element it is directly connected to. This transmission element
(e.g., a photonic cross-connect) then interfaces with the DWDM
systems (if present) and so forth. This allows for the photonic
switches to discover the transmission network topology and
characteristics prior to attached devices asking for connections. It
also caters for the continued support of any proprietary signaling
that may already exist between DWDM and/or other transmission
systems (whether in-band or out-of-band). All that is required is
support of the standard external signaling interface.
We also propose these signaling flows be supported on a dedicated
wavelength, configured throughout the network. Whether this
wavelength is part of the standard ITU grid or not, is beyond the
scope of this document. We recommend however, that the signaling
wavelength be a standard ITU channel, considering that the
combination of existing C-band (1530- to 1560-nm) and the emerging
S- (upper 1400-nm region) and L- (1570- to 1625-nm) transmission
bands will leave little room for suitable non-ITU wavelengths.
Ceuppens/Blumenthal et al. [Page 9]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
Since dedicating an entire wavelength might not always be viable, we
should envision the possibility of using this wavelength also for
data traffic and envisage a way of sending the non-time-critical
traffic in between the management traffic.
The signaling protocol can easily be based on existing protocols. A
slightly modified OSPF can be used for optical network topology
discovery and distribution, as well as for route computation and
path selection. Topology advertisement includes not only the nodes
and the links to the nodes, but also characteristics of the links.
The actual signaling protocol can be RSVP as extended for MPLS/TE.
Finally, path management includes monitoring the path for failures,
knowledge of failure restoration policies, and path teardown.
6. Summary
This document outlined a proposal to expand Multi-Protocol Lambda
Switching in two areas:
- Include network elements such as DWDM, OADM and ATM switches to
create a versatile transparent and open photonic network.
- Expand the work beyond basic connection establishment and include
performance-monitoring capabilities
Work in this area should be closely coordinated with activities in
the T1 committee, ITU and OIF to ensure a consistent industry-wide
solution.
8. Security Considerations
This document raises no new security issues.
9. References
[1] Awduche, D., Y. Rekhter, J. Drake and R. Coltun, "Multi-Protocol
Lambda Switching: Combining MPLS Traffic Engineering Control
With Optical Crossconnects", work in progress, November 1999.
[2] Awduche, D., J. Malcolm, J. Agogbua, M. O'Dell and J. McManus,
"Requirements for Traffic Engineering Over MPLS", RFC 2702,
September 1999.
10.Acknowlegdements
We would like to thank Curtis Brownmiller (MCI WorldCom) for his
review and comments.
Ceuppens/Blumenthal et al. [Page 10]
Internet Draft draft-ceuppens-mpls-optical-00.txt March 2000
11.Author's Addresses
Luc Ceuppens W.L. Edwards
Chromisys iLambda Networks
1012 Stewart Drive Aspen, CO
Sunnyvale, CA 94086 970.948.7104
Email: lceuppens@Chromisys.com Email: texas@ilambda.com
Dan Blumenthal Jacek Chrostowski
Chromisys Cisco Systems
421 Pine Avenue 365 March Rd
Goleta, CA 93117 Canata, Ontario K2K2C9
Email: danb@Chromisys.com Email: jchrosto@cisco.com
John Drake
Chromisys
1012 Stewart Drive
Sunnyvale, CA 94086
Email: jdrake@Chromisys.com
Ceuppens/Blumenthal et al. [Page 11]