-----Original Message-----
From: Thomas D. Nadeau [mailto:[EMAIL PROTECTED]
Sent: Sunday, November 27, 2005 6:53 PM
To: Black, David
Cc: [email protected]; [EMAIL PROTECTED]; [EMAIL PROTECTED];
[EMAIL PROTECTED]; [EMAIL PROTECTED]; [EMAIL PROTECTED];
[EMAIL PROTECTED]
Subject: Re: Gen-ART review of MPLS OAM Requirements and
Framework Drafts
Importance: High
This response addresses the comments directed at
the OAM Requirements draft. Please see inline for responses
(or questions if necessary).
-- draft-ietf-mpls-oam-requirements-06.txt
I found a significant number of typos and English language
editorial cleanup nits that I'll leave to the RFC Editor to
address. It might be good to alert the RFC Editor that this
draft is likely to require heavy editorial attention.
Section 2.1 Terminology - Both "Head-end Label Switch
Router (LSR)" and "Head-end Label Switching Router (LSR)"
are defined. Only one of these two definitions should be
used - the other should be removed.
Section 2.2 Acronyms -
LSP: Label Switch Path
LSR: Label Switch Router
I suspect those should be "Label Switched Path" and "Label
Switching Router". In any case, between this and the previous
comment, consistency on whether "LS" in acronyms stands for
"Label Switch" vs. "Label Switched" and/or "Label Switching"
is needed throughout the draft.
Changed LSR to "Label Switching Router" and LSP to
"Label Switched Path" as per RFC3031 and also globally
made things consistent throughout the document.
Section 3 Motivations - RFC 2119 terminology is not appropriate
in this section, as it does not express requirements. Change
all occurrences of "MAY" to "may".
Fixed: MAY -> may
The statement that:
However as of this writing, existing drafts focus on single
provider solutions or focus on a single aspect of the MPLS
architecture or application of MPLS.
appears to be inconsistent with Section 4.9's discussion of
standard MPLS MIBs, which includes references to the RFCs that
specify them. In addition, this statement will be dramatically
incorrect when the LSP Ping draft is published. It should be
rewritten to be more about the drafts that respond to these
requirements rather than a criticism of what has gone before.
I think this is short and sweet and no longer has
the conflicts you mentioned:
3. Motivations
This document was created in order to provide requirements
which could be used to create consistent and useful OAM
functionality that meets operational requirements of those
service providers who have or are deploying MPLS.
Section 4.1 Detection of Label Switched Path Defects
The characterization of "reasonable amount of time" for
detection as "both predictable and consistent" is incomplete.
For example, always detecting defects 47-48 hours after they
manifest in customer traffic is "both predictable and
consistent", but not "reasonable".
That is true; however, the discussion of how "reasonable"
or "acceptable" is defined, is really based on the provider and
the SLA offered to the customer in question.
A ballpark/approximate
order-of-magnitude discussion should be added covering the
time frames in which LSP defects will likely affect customer
traffic, and the detection timeframes required for operator
detection to precede customers noticing that something is
wrong.
I hope this makes things clearer. In Particular, see the last
two sentences:
4.1 Detection of Label Switched Path Defects
The ability to detect defects in a broken Label Switched Path
(LSP) SHOULD not require manual hop-by-hop troubleshooting of
each LSR used to switch traffic for that LSP. For example,
it is not desirable to manually visit each LSR along the data
plane path used to transport an LSP; instead,this function
SHOULD be automated and able to be performed at some operator
specified frequency from the origination point of that LSP.
This implies solutions that are interoperable as to allow for
such automatic operation. Furthermore, the automation of path
liveliness is desired in cases where large numbers of LSPs might
be tested. For example, automated ingress LSR to egress
LSR testing
functionality is desired for some LSPs. The goal is to detect LSP
path defects before customers do, and this requires detection and
correction of LSP defects in a manner that is both predictable
and
sufficiently within the constraints of the service level
agreement
under which the service is being offered. Simply put, the sum of
the time it takes an OAM tool to detect a defect and
the time needed for an operational support system to react to
this defect by possibly correcting it or notifying the customer,
must fall within the bounds of the service level agreement in
question.
The discussion of ICMP-based ping appears to assume that
ICMP-based ping is an MPLS control plane function, even at
LSRs that only forward an ICMP packet (i.e., not the destination
of the ICMP packet, and not the node at which the TTL hits zero).
If that assumption is correct, it needs to be stated, but it
does not appear to be consistent with RFC 3032's discussion
of ICMP and MPLS, which specifically states that ICMP packets
can be label switched. If ICMP-based ping goes through the
data plane as long as the TTL has not expired (which is what
I would expect), then this paragraph needs to be reworked.
ICMP-based ping packets are carried within MPLS LSPs
and so are label-switched, but that is not guaranteed. Thus
inconsistencies can result (and be unpredictable). Here is the
revised text:
Although ICMP-based ping [RFC792] can be sent through an LSP as
an IP payload, the use of this tool to verify the defect
free operation
of an LSP has the potential for returning erroneous results (both
positive and negative) for a number of reasons. First, since
the ICMP traffic is based on legally addressable IP addressing,
it is possible for ICMP messages that are originally transmitted
inside of an LSP to "fall out of the LSP" at some point along
the path. In these cases, since ICMP packets are routable
a falsely positive response may be returned. In other cases
where the data plane of a specific LSP needs to be tested, it
is difficult to guarantee that traffic based on an ICMP ping
header is parsed and hashed to the same equal-cost multi-paths
as the data traffic.
Any detection mechanisms that depend on receiving status via a
return path SHOULD provide multiple return options with the
expectation that one of them will not be impacted by the original
defect. An example of a case where a false negative might occur
would be a mechanism that requires a functional MPLS return path.
Since MPLS LSPs are unidirectional, it is possible that although
the forward LSP which is the LSP under test, might be
functioning,
the response from the destination LSR might be lost, thus giving
the source LSR the false impression that the forward LSP is
defective. However, if an alternate return path could be
specified -- say IP for example -- then the source could
specify this as the return path to the destination, and in
this case, would receive a response indicating to it that
the return LSP is defective.
Section 4.4 Service Level Agreement Measurement
... the qualitative
aspects of LSPs such as jitter, delay, latency and loss
I believe "quantitative" should be substituted for "qualitative".
Done.
Section 4.5 Frequency of OAM Execution
... probe-based detection mechanisms
that incur significant mismatches in the probe rate
That doesn't parse for me - what is mismatched to what? If
it's the rates of probing by multiple probe-based mechanisms,
the text quoted above should read:
... probe frequency or rate differences among multiple
probe-based detection mechanisms
How about:
This includes the frequency of the execution of any OAM
functions. The capability to synchronize OAM operations
is required
as to to permit consistent measurement of service level
agreements.
To elaborate, there are defect conditions such as misbranching or
misdirection of traffic for which probe-based detection
mechanisms
that incur significant mismatches in their detection frequency
MAY
result in flapping. This can be addressed either by synchronizing
the rate or having the probes self-identify their probe rate.
For example, when the probing mechanisms are bootstrapping,
they might negotiate and ultimately agree on a probing rate,
therefore providing a consistent probing frequency and avoiding
the aforementioned problems.
At the end of the section, I understand how synchronizing
probe rates helps address the problem of inconsistent views of
network problem state, but explanation is needed for how
"having the probes self-identify their probe rate" helps.
Gotcha. See last sentence above.
Section 4.11 Per-LSP Accounting Requirements - ASBR acronym
used without expansion or definition.
Added ASBR to acronyms section.
Section 4.11.1 Requirements - The use of the word "key" in
(4) invites confusion with security usage of that word.
Please pick another term such as "tag" or "identifier",
unless there is a specific need to use "key".
(4) All LSRs that contain LSPs that are being measuremented
need to have a common identifier to distinguish each LSP.
The identifier MUST be unique to each LSP, and its
mapping to
LSP SHOULD be provided from whether manual or automatic
configuration.
Section 4.11.2 Scalability - This is not about "scalability"
per se, but rather minimum accounting requirements motivated
by scaling considerations. It should be retitled to "Location
of Accounting" or something like that.
Section 5 Security Considerations
The term "tools" in the first paragraph should be changed
to "network mechanisms", which is I believe this paragraph
is primarily about (e.g., authorization restrictions on use
of LSP Ping); tools is susceptible to being misread as being
primarily about the software applications that inspect the
network as opposed to the network mechanisms they use.
Done. Here is the new text:
5. Security Considerations
Provisions to any of the network mechanisms designed to satisfy
the requirements described herin are required to prevent their
unauthorized use. Likewise, these network mechanisms MUST
provide a means by which an operator can prevent denial of
service attacks if those network mechanisms are used in such
an attack.
LSP mis-merging has security implications beyond that of simply
being a network defect. LSP mis-merging can happen due to a
number
of potential sources of failure, some of which (due to MPLS label
stacking) are new to MPLS.
The performance of diagnostic functions and path characterization
involve extracting a significant amount of information about
network construction which the network operator MAY consider
private.
Sections 6 and 11 on IANA Considerations are duplicative
(but consistent). Remove one of them.
Done; removed last one.
I am attaching the revised version w/o formatting.
I will send this version to the I-D editor with formatting
as soon as I hear back from you that these changes are okay
(hopefully tomorrow -- Monday).
--Tom
Network Working Group Thomas D. Nadeau
Internet Draft Monique Morrow
Expires: June 2006 George Swallow
Cisco Systems, Inc.
David Allan
Nortel Networks
Satoru Matsushima
Japan Telecom
December 2005
OAM Requirements for MPLS Networks
draft-ietf-mpls-oam-requirements-07.txt
Status of this Memo
By submitting this Internet-Draft, each author represents that
any applicable patent or other IPR claims of which he or she is
aware have been or will be disclosed, and any of which he or she
becomes aware will be disclosed, in accordance with Section 6 of
BCP 79.
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/1id-abstracts.html
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
Abstract
As transport of diverse traffic types such as voice, frame
relay, and ATM over MPLS become more common, the ability
to detect,
handle and diagnose control and data plane defects becomes
critical.
Detection and specification of how to handle those defects is not
only important because such defects may not only affect the
fundamental operation of an Multi-Protocol Label Switching (MPLS)
network, but also because they MAY impact service level
specification
commitments for customers of that network.
This document describes requirements for user and data
plane operations and management for MPLS.
These requirements have been gathered
from network operators who have extensive experience deploying
MPLS networks, similarly some of these
requirements have appeared in other documents. This draft
specifies
Operations and Management requirements for Multi-Protocol Label
Switching, as well as for applications of Multi-Protocol Label
Switching such as pseudowire voice and virtual private network
services. Those interested in specific issues relating to
instrumenting MPLS for Operations
and Management purposes are directed to the Multi-Protocol Label
Switching Architecture specification.
Table of Contents
Abstract.....................................................1
1. Introduction.................................................2
2. Document Conventions.........................................2
2.1 Terminology..................................................2
2.2 Acronyms.....................................................2
3. Motivations..................................................2
4. Requirements.................................................2
5. Security Considerations.....................................26
6. IANA considerations.........................................27
7. References..................................................27
7.1 Normative references........................................27
7.2 Informative references......................................29
8. Author's Addresses..........................................29
9. Intellectual Property Notice................................30
10. Full Copyright Statement....................................29
1. Introduction
This document describes requirements for user and data
plane operations and management (OAM) for Multi-Protocol
Label Switching (MPLS). These requirements have been gathered
from network operators who have extensive experience deploying
MPLS networks. This draft specifies OAM requirements
for MPLS, as well as for applications of MPLS.
No specific mechanisms are proposed to address these
requirements at this time. The goal of this draft is to
identify a commonly applicable set of requirements for MPLS
OAM at this time. Specifically, a set of requirements that apply
to the most common set of MPLS networks deployed by service
provider organizations today. These requirements can then be used
as a base for network management tool development and to guide
the evolution of currently specified tools, as well as the
specification of OAM functions that are intrinsic to protocols
used in MPLS networks.
Comments should be made directly to the MPLS mailing list
at [EMAIL PROTECTED]
2. Document Conventions
2.1 Terminology
The key words "MUST", "MUST NOT", "REQUIRED", "SHALL",
"SHALL NOT",
"SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and
"OPTIONAL" in this
document are to be interpreted as described in RFC 2119.
Defect: Any error condition that prevents an Label Switched
Path functioning correctly. For example, loss of an
IGP path will most likely also result in an Label
Switched Path not being able to deliver traffic to its
destination. Another example is the breakage of
a TE tunnel. These MAY be due to physical
circuit failures or failure of switching nodes
to operate as expected.
Multi-vendor/multi-provider network operation typically
requires agreed upon definitions of defects (when it is
broken and when it is not) such that both recovery
procedures and service level specification impacts can
be specified.
Probe-based-detection: Active measurement using a tool such as
Label Switched Path Ping.
Collecting traffic: Passive measurement of network traffic.
Head-end Label Switching Router (LSR): The beginning of a label
switched path. A head-end LSR is also referred to as
an Ingress Label Switching Router.
Probe-based-detection: Active measurement using a tool such as
Label Switched Path Ping.
Collecting traffic: Passive measurement of network traffic.
propagation latency: delay added by the propagation of the packet
through the link (fixed value that
depends on
the distance of the link and the propagation
speed).
transmission latency: delay added by the transmission of
the packey
over the link i.e. the time it takes put
the
packet over the media (value that
depends of
the link throughput and packet length).
processing latency: delay added by all the operations
related to the
switching of labeled packet (value is node
implementation specific and may are
considered
as fixed and constant for a given equipment).
queuing/buffering latency: delay caused by packet queuing
(value is
variable since depending on the packet
arrival rate in addition to the
dependance on the packet length and
the
link throughput).
node latency: delay added by the network element resulting
from of
the sum of the transmission, processing and
queuing/
buffering latency
one-hop delay: fixed delay experienced by a packet to
reach the next
hop reesulting from the of the propagation
latency,
the transmission latency and the
processing latency.
minimum path latency: sum of the one-hop delays experienced by
the
packet when travelling from the
ingress to the
egress LSR.
variable path latency (jitter): sum of the delays caused by the
queuing latency experienced by
the
packet at each node over the
path.
2.2 Acronyms
ASBR: Autonomous System Border Router
CE: Customer Edge
SP: Service Provider
ECMP: Equal Cost Multipath
LSP: Label Switched Path
LSP Ping: Label Switched Path Ping
LSR: Label Switching Router
OAM: Operations and Management
RSVP: Resource reSerVation Protocol
LDP: Label Distribution Protocol
DoS: Denial of service
3. Motivations
This document was created in order to provide requirements
which could be used to create consistent and useful OAM
functionality that meets operational requirements of those
service providers who have or are deploying MPLS.
4. Requirements
The following sections enumerate the OAM requirements
gathered from service providers who have deployed MPLS
and services based on MPLS networks. Each requirement is
specified in detail to clarify its applicability.
Although the requirements specified herein are defined by
the IETF, they have been harmonized with requirements
gathered by other standards bodies such as the ITU [Y1710].
4.1 Detection of Label Switched Path Defects
The ability to detect defects in a broken Label Switched Path
(LSP) SHOULD not require manual hop-by-hop troubleshooting of
each LSR used to switch traffic for that LSP. For example,
it is not desirable to manually visit each LSR along the data
plane path used to transport an LSP; instead,this function
SHOULD be automated and able to be performed at some operator
specified frequency from the origination point of that LSP.
This implies solutions that are interoperable as to allow for
such automatic operation. Furthermore, the automation of path
liveliness is desired in cases where large numbers of LSPs might
be tested. For example, automated ingress LSR to egress
LSR testing
functionality is desired for some LSPs. The goal is to detect LSP
path defects before customers do, and this requires detection and
correction of LSP defects in a manner that is both predictable
and
sufficiently within the constraints of the service level
agreement
under which the service is being offered. Simply put, the sum of
the time it takes an OAM tool to detect a defect and
the time needed for an operational support system to react to
this defect by possibly correcting it or notifying the customer,
must fall within the bounds of the service level agreement in
question.
Synchronization of detection time bounds by tools used to detect
broken LSPs is required. Failure to specifying defect detection
time bounds may result in an ambiguity in test results. If the
time to detect is known, then automated responses can be
specified
both with respect to and with regard to resiliency and service
level specification reporting. Further, if synchronization of
detection time bounds is possible, an operational framework
can be
established that can guide the design and specification of MPLS
applications.
Although ICMP-based ping [RFC792] can be sent through an LSP as
an IP payload, the use of this tool to verify the defect free
operation
of an LSP has the potential for returning erroneous results (both
positive and negative) for a number of reasons. First, since
the ICMP traffic is based on legally addressable IP addressing,
it is possible for ICMP messages that are originally transmitted
inside of an LSP to "fall out of the LSP" at some point along
the path. In these cases, since ICMP packets are routable
a falsely positive response may be returned. In other cases
where the data plane of a specific LSP needs to be tested, it
is difficult to guarantee that traffic based on an ICMP ping
header is parsed and hashed to the same equal-cost multi-paths
as the data traffic.
Any detection mechanisms that depend on receiving status via a
return path SHOULD provide multiple return options with the
expectation that one of them will not be impacted by the original
defect. An example of a case where a false negative might occur
would be a mechanism that requires a functional MPLS return path.
Since MPLS LSPs are unidirectional, it is possible that although
the forward LSP which is the LSP under test, might be
functioning,
the response from the destination LSR might be lost, thus giving
the source LSR the false impression that the forward LSP is
defective. However, if an alternate return path could be
specified -- say IP for example -- then the source could
specify this as the return path to the destination, and in
this case, would receive a response indicating to it that
the return LSP is defective.
The OAM packet MUST follow exactly the customer data path in
order
to reflect path liveliness used by customer data. Particular
cases
of interest are forwarding mechanisms such as equal cost
multipath
(ECMP) scenarios within the operator's network whereby flows are
load-shared across parallel (i.e., equal IGP cost) paths. Where
the customer traffic MAY be spread over multiple paths, what is
required is to be able to detect failures on any of the path
permutations. Where the spreading mechanism is payload specific,
payloads need to have forwarding that is common with the traffic
under test. Satisfying these requirements introduces complexity
into ensuring that ECMP connectivity permutations are exercised,
and that defect detection occurs in a reasonable amount of time.
4.2 Diagnosis of a Broken Label Switched Path
The ability to diagnose a broken LSP and to isolate the failed
component (i.e., link or node) in the path is required. For
example, note that specifying recovery actions for misbranching
defects in an LDP network is a particularly difficult case.
Diagnosis of defects and isolation of the failed component is
best accomplished via a path trace function which can return the
the entire list of LSRs and links used by a certain LSP (or at
least the set of LSRs/links up to the location of the defect) is
required. The tracing capability SHOULD include the ability to
trace recursive paths, such as when nested LSPs are used. This
path trace function MUST also be capable of diagnosing LSP
mis-merging by permitting comparison of expected vs. actual
forwarding behavior at any LSR in the path. The path trace
capability SHOULD be capable of being executed from both the
head-end Label Switching Router (LSR) and MAY permit downstream
path components to be traced from an intermediate mid-point LSR.
Additionally, the path trace function MUST have the ability to
support equal cost multipath scenarios described above in
section 4.1.
4.3 Path characterization
The path characterization function is the ability to
reveal details
of LSR forwarding operations. These details can then be compared
later during subsequent testing relevant to OAM functionality.
This would include but is not limited to:
- consistent use of pipe or uniform time to live (TTL)
models by
an LSR [RFC3443].
- sufficient details that allow the test origin to
excersize all path permutations related to load spreading
(e.g. ECMP).
- stack operations performed by the LSR, such as pushes, pops,
and TTL propagation at penultimate hop LSRs.
4.4 Service Level Agreement Measurement
Mechanisms are required to measure the diverse aspects of Service
Level Agreements which include:
- defect free forwarding. The service is considered to be
available and the other aspects of performance measurement
listed below have meaning, or the service is unavailable and
other aspects of performance measurement do not.
- latency - amount of time required for traffic to transit
the network
- packet loss
- jitter - measurement of latency variation
Such measurements can be made independently of the user traffic
or via a hybrid of user traffic measurement and OAM probing.
At least one mechanism is required to measure the number
of OAM packets. In addition, the ability to measure the
quantitative aspects of LSPs such as jitter, delay, latency and
loss MUST be available in order to determine whether or not the
traffic for a specific LSP are traveling within the
operator-specified tollerances.
Any method considered SHOULD be capable of measuring the latency
of an LSP with minimal impact on network resources. See section
2.1 for definitions of the various quantitative aspects of LSPs.
4.5 Frequency of OAM Execution
The operator MUST have the flexibility to configure OAM
parameters insofaras to meet their specific operational
requirements.
This includes the frequency of the execution of any OAM
functions. The capability to synchronize OAM operations
is required
as to to permit consistent measurement of service level
agreements.
To elaborate, there are defect conditions such as misbranching or
misdirection of traffic for which probe-based detection
mechanisms
that incur significant mismatches in their detection frequency
MAY
result in flapping. This can be addressed either by synchronizing
the rate or having the probes self-identify their probe rate.
For example, when the probing mechanisms are bootstrapping,
they might negotiate and ultimately agree on a probing rate,
therefore providing a consistent probing frequency and avoiding
the aforementioned problems.
One observation would be that wide-spread deployment of
MPLS, common
implementation of monitoring tools and the need for
inter-carrier synchronization of defect and service level
specification handling will drive specification of OAM parameters
to commonly agreed on values and such values will have to be
harmonized with the surrounding technologies (e.g. SONET/SDH,
ATM etc.) in order to be useful. This will become particularly
important as networks scale and misconfiguration can result in
churn, alarm flapping etc.
4.6 Alarm Suppression, Aggregation and Layer Coordination
Network elements MUST provide alarm suppression functionality
that
prevents the generation of superfluous generation of alarms by
simply discarding them (or not generating them in the
first place),
or by aggregating them together, and thereby greatly reducing the
number of notifications emitted. When viewed in conjuction with
requirement 4.7 below, this typically requires fault notification
to the LSP egress that
MAY have specific time constraints if the application
using the LSP
independently implements path continuity testing (for example ATM
I.610 Continuity check (CC)[I610]). These constraints apply to
LSPs that are monitored. The nature of MPLS applications allows
for the possibility to have multiple MPLS applications attempt to
respond to defects simultaneously. For example, layer-3 MPLS VPNs
that utilize Traffic Engineered tunnels, where a failure
occurs on
the LSP carrying the Traffic Engineered tunnel. This failure
would
affect he VPN traffic that uses the tunnel's LSP. Mechanisms are
required to coordinate network response to defects.
4.7 Support for OAM Interworking for Fault Notification
An LSR supporting the interworking of one or more networking
technologies over MPLS MUST be able to translate an MPLS defect
into the native technology's error condition. For example, errors
occurring over a MPLS transport LSP that supports an emulated
ATM VC MUST translate errors into native ATM OAM Alarm Indication
Signal (AIS) cells at the termination points of the LSP. The
mechanism SHOULD consider possible bounded detection time
parameters, e.g., a "hold off" function before reacting as to
synchronize with the OAM functions.
One goal would be alarm suppression by the upper layer using
the LSP. As observed in section 4.5, this requires that MPLS
perform detection in a bounded timeframe in order to initiate
alarm suppression prior to the upper layer independently
detecting the defect.
4.8 Error Detection and Recovery.
Recovery from a fault by a network element can be facilitated by
MPLS OAM procedures. These procesures will detect a broader range
of defects than that of simple link and node failures.
Since MPLS LSPs may span multiple routing areas and
service provider
domains, fault recovery and error detection should be possible
in these configuration as well as in the more simplifed
single-area/domain configurations.
Recovery from faults SHOULD be automatic. It is a requirement
that
faults SHOULD be detected (and possibly corrected) by the network
operator prior to customers of the service in question detecting
them.
4.9 Standard Management Interfaces
The wide-spread deployment of MPLS requires common information
modeling of management and control of OAM functionality. This is
reflected in the the integration of standard MPLS-related MIBs
(e.g. [RFC3813][RFC3812][RFC3814]) for fault, statistics and
configuration management. These standard interfaces provide
operators with common programmatic interface access to
operations and management functions and their status.
4.10 Detection of Denial of Service Attacks
The ability to detect denial of service (DoS) attacks against the
data or control planes MUST be part of any security management
related to MPLS OAM tools or techniques.
4.11 Per-LSP Accounting Requirements
In an MPLS network, service providers (SPs) can measure traffic
from an LSR to the egress of the network using some MPLS related
MIBs, for example. This means that it is a reasonable to know how
much traffic is traveling from where to where (i.e., a traffic
matrix) by analyzing the flow of traffic. Therefore, traffic
accounting in an MPLS network can be summarized as the following
three items.
(1) Collecting information to design network
Providers and their customers MAY need to verify high-level
service level specifications, either to continuously
optimize their networks, or to offer guaranteed bandwidth
services. Therefore, traffic accounting to monitor MPLS
applications is required.
(2) Providing a Service Level Specification
For the purpose of optimized network design, a service
provider may offer the traffic informationr. Optimizing
network design needs this information.
(3) Inter-AS environment
Service providers that offer inter-AS services require
accounting of those services.
These three motivations need to satisfy the following.
- In (1) and (2), collection of information on a per-LSP
basis is a minimum level of granularity of collecting
accounting information at both of ingress and egress
of an LSP.
- In (3), SP's ASBR carry out interconnection functions
as an
intermediate LSR. Therefore, identifying a pair of ingress
and egress LSRs using each LSP is needed to determine the
cost of the service that a customer is using.
4.11.1 Requirements
Accounting on a per-LSP basis encompasses the following set of
functions:
(1) At an ingress LSR accounting of traffic through LSPs
beginning at each egress in question.
(2) At an intermediate LSR, accounting of traffic through
LSPs for each pair of ingress to egress.
(3) At egress LSR, accounting of traffic through LSPs
for each ingress.
(4) All LSRs that contain LSPs that are being measuremented
need to have a common identifier to distinguish each LSP.
The identifier MUST be unique to each LSP, and its
mapping to
LSP SHOULD be provided from whether manual or automatic
configuration.
In the case of non-merged LSPs, this can be achieved by
simply reading traffic counters for the label stack
associated
with the LSP at any LSR along its path. However, in order to
measure merged LSPs, an LSR MUST have a means to distinguish
the source of each flow so as to disambiguate the statistics.
4.11.2 Scalability
It is not realistic to perform the just described operations by
LSRs in a network on all LSPs that exist in a network.
At a minimum, per-LSP based accounting SHOULD be
performed on the
edges of the network -- at the edges of both LSPs and the MPLS
domain.
5. Security Considerations
Provisions to any of the network mechanisms designed to satisfy
the requirements described herin are required to prevent their
unauthorized use. Likewise, these network mechanisms MUST
provide a means by which an operator can prevent denial of
service attacks if those network mechanisms are used in such
an attack.
LSP mis-merging has security implications beyond that of simply
being a network defect. LSP mis-merging can happen due to a
number
of potential sources of failure, some of which (due to MPLS label
stacking) are new to MPLS.
The performance of diagnostic functions and path characterization
involve extracting a significant amount of information about
network construction which the network operator MAY consider
private.
6. IANA Considerations
This document creates no new requirements on IANA namespaces
[RFC2434].
7. References
7.1 Informative References
[RFC3812] Srinivasan, C., Viswanathan, A. and T.
Nadeau, "MPLS Traffic Engineering Management
Information Base Using SMIv2", RFC3812, June 2004.
[RFC3813] Srinivasan, C., Viswanathan, A. and T.
Nadeau, "MPLS Label Switch Router Management
Information Base Using SMIv2", RFC3813, June 2004.
[RFC3814] Nadeau, T., Srinivasan, C., and A.
Viswanathan, "Multiprotocol Label Switching
(MPLS) FEC-To-NHLFE (FTN) Management
Information Base", RFC3814, June 2004.
[Y1710] ITU-T Recommendation Y.1710, "Requirements for
OAM Functionality In MPLS Networks"
[I610] ITU-T Recommendation I.610, "B-ISDN operations and
maintenance principles and functions", February 1999
[RFC2434] Narten, T. and H. Alvestrand, "Guidelines for Writing
an IANA Considerations section in RFCs", BCP 26, RFC
2434, October 1998.
[RFC792] Postel, J., "Internet Control Message Protocol",
RFC792,
September 1981.
[RFC3443] Agarwal, P, Akyol, B., "Time To Live (TTL)
Processing in
Multi-Protocol Label Switching (MPLS)
Networks.", RFC3443,
January 2003.
8. Authors' Addresses
Thomas D. Nadeau
Cisco Systems, Inc.
300 Beaver Brook Road
Boxboro, MA 01719
Phone: +1-978-936-1470
Email: [EMAIL PROTECTED]
Monique Jeanne Morrow
Cisco Systems, Inc.
Glatt-Com, 2nd Floor
CH-8301
Switzerland
Voice: (0)1 878-9412
Email: [EMAIL PROTECTED]
George Swallow
Cisco Systems, Inc.
300 Beaver Brook Road
Boxboro, MA 01719
Voice: +1-978-936-1398
Email: [EMAIL PROTECTED]
David Allan
Nortel Networks
3500 Carling Ave.
Ottawa, Ontario, CANADA
Voice: 1-613-763-6362
Email: [EMAIL PROTECTED]
Satoru Matsushima
Japan Telecom
4-7-1, Hatchobori, Chuo-ku
Tokyo, 104-8508 Japan
Phone: +81-3-5540-8214
Email: [EMAIL PROTECTED]
9. Intellectual Property Statement
The IETF takes no position regarding the validity or scope of any
Intellectual Property Rights or other rights that might be
claimed to
pertain to the implementation or use of the technology
described in
this document or the extent to which any license under such
rights
might or might not be available; nor does it represent that it
has
made any independent effort to identify any such rights.
Information
on the procedures with respect to rights in RFC documents can be
found in BCP 78 and BCP 79.
Copies of IPR disclosures made to the IETF Secretariat and any
assurances of licenses to be made available, or the result of an
attempt made to obtain a general license or permission for the
use of
such proprietary rights by implementers or users of this
specification can be obtained from the IETF on-line IPR
repository at
http://www.ietf.org/ipr.
The IETF invites any interested party to bring to its
attention any
copyrights, patents or patent applications, or other proprietary
rights that may cover technology that may be required to
implement
this standard. Please address the information to the
IETF at ietf-
[EMAIL PROTECTED]
10. Full Copyright Statement
Copyright (C) The Internet Society (2005). This document is
subject to the rights, licenses and restrictions contained in
BCP
78, and except as set forth therein, the authors retain all
their
rights.
This document and the information contained herein are provided
on an "AS IS" basis and THE CONTRIBUTOR, THE ORGANIZATION HE/SHE
REPRESENTS OR IS SPONSORED BY (IF ANY), THE INTERNET SOCIETY AND
THE INTERNET ENGINEERING TASK FORCE DISCLAIM ALL WARRANTIES,
EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO ANY
WARRANTY THAT
THE USE OF THE INFORMATION HEREIN WILL NOT INFRINGE ANY
RIGHTS OR
ANY IMPLIED WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A
PARTICULAR PURPOSE.
11. Acknowledgment
Funding for the RFC Editor function is currently provided by the
Internet Society.
The authors wish to acknowledge and thank the following
individuals for their valuable comments to this document:
Adrian Smith, British Telecom; Chou Lan Pok, SBC; Mr.
Ikejiri, NTT Communications and Mr.Kumaki of KDDI.
Hari Rakotoranto, Miya Khono, Cisco Systems; Luyuan Fang, AT&T;
Danny McPherson, TCB; Dr.Ken Nagami, Ikuo Nakagawa, Intec
Netcore,
and David Meyer.