Network Working Group J. Schoenwaelder
Internet-Draft Jacobs University Bremen
Intended status: Informational June 15, 2007
Expires: December 17, 2007
Lessons Learned from the SMIng Project
draft-schoenw-sming-lessons-00.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/ietf/1id-abstracts.txt.
The list of Internet-Draft Shadow Directories can be accessed at
http://www.ietf.org/shadow.html.
This Internet-Draft will expire on December 17, 2007.
Copyright Notice
Copyright (C) The IETF Trust (2007).
Abstract
A data modeling language for network management protocols called
SMIng was developed within the Network Management Research Group of
the Internet Research Task Force over a period of several years.
This memo documents some of the lessons learned during the project
for consideration by designers of future data modeling languages for
network management protocols.
Schoenwaelder Expires December 17, 2007 [Page 1]
Internet-Draft Lessons Learned from the SMIng Project June 2007
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3
2. SMIng Goals and History . . . . . . . . . . . . . . . . . . . 3
3. Lessons Learned . . . . . . . . . . . . . . . . . . . . . . . 4
3.1. Readers, Writers, Tool Developers . . . . . . . . . . . . 4
3.2. Data vs. Command vs. Object vs. Document . . . . . . . . . 5
3.3. Naming Independence . . . . . . . . . . . . . . . . . . . 5
3.4. Errors and Exceptions . . . . . . . . . . . . . . . . . . 6
3.5. Configuration Storage Models . . . . . . . . . . . . . . . 6
3.6. Selection of Language Features . . . . . . . . . . . . . . 7
3.7. Syntax versus Semantics . . . . . . . . . . . . . . . . . 7
3.8. Reuse of Definitions . . . . . . . . . . . . . . . . . . . 7
3.9. Versioning and Meta Information . . . . . . . . . . . . . 8
3.10. Extensibility . . . . . . . . . . . . . . . . . . . . . . 8
3.11. Language Independence . . . . . . . . . . . . . . . . . . 8
3.12. Compliance and Conformance . . . . . . . . . . . . . . . . 9
4. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . 9
5. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 9
6. Security Considerations . . . . . . . . . . . . . . . . . . . 10
7. Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . 10
8. Informative References . . . . . . . . . . . . . . . . . . . . 10
Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 10
Intellectual Property and Copyright Statements . . . . . . . . . . 12
Schoenwaelder Expires December 17, 2007 [Page 2]
Internet-Draft Lessons Learned from the SMIng Project June 2007
1. Introduction
The Network Management Research Group (NMRG) of the Internet Research
Task Force (IRTF) has developed a new data modeling language for
network management protocols called SMIng [RFC3780]. Work on SMIng
started in the late 1999 and concluded with the publication of the
core language [RFC3780] and an extension which provides the mappings
to the Simple Network Management Protocol (SNMP) [RFC3781] in 2004.
This memo documents some of the lessons learned during the project
for consideration by engineers of future data modeling languages for
network management protocols. Our focus is on technical design
aspects that have turned out to be more difficult than expected as
well as aspects that required significant discussions in order to
understand trade-offs.
2. SMIng Goals and History
SMIng started in 1999 as a research project to address some drawbacks
of SMIv2, the current data modeling language for management
information bases. Primarily, its partial dependence on ASN.1 and a
number of exception rules turned out to be problematic. In 2000, the
work was handed over to the IRTF Network Management Research Group
where it was significantly detailed. Since the work of the RAP
Working Group on COPS-PR and SPPI emerged in 1999/2000, SMIng was
split into two parts: a core data definition language (defined in
this document) and protocol mappings to allow the application of core
definitions through (potentially) multiple management protocols. The
replacement of SMIv2 and SPPI by a single merged data definition
language was also a primary goal of the IETF SMING Working Group that
was chartered at the end of 2000. The SMING Working Group could not
reach agreement, resulting in the Working Group being closed down in
April 2003. The NMRG then published the SMIng specifications in
[RFC3780] and [RFC3781] in order to document the results achieved in
the NMRG.
In short, the published version of SMIng aimed at being a data
modeling language that can be adapted for different management
protocols and thus be closer to an information modeling language.
For a more detailed discussion of the terms data modeling language
and information modeling languages, we refer the reader to RFC 3444
[RFC3444].
The problem which drove the development of SMIng, namely duplicated
data modeling efforts and often also duplicated instrumentation code,
has not been solved and it even may have become more pressing. A
data modeling language capable to drive tools that can automate the
Schoenwaelder Expires December 17, 2007 [Page 3]
Internet-Draft Lessons Learned from the SMIng Project June 2007
implementation of command line interfaces, monitoring interfaces, and
configuration interfaces from a single specification surely would
have significant value.
During the SMIng project, people have looked at various issues that
have to be dealt with in this context. The purpose of this memo is
to document these issues and some of the alternatives that have been
explored to tackle them so that future language designers can learn
from the SMIng experiment.
3. Lessons Learned
This section summarizes the lessons learned from the SMIng project.
Designers of future data modeling languages for network management
protocols are encouraged to study these lessons to avoid potential
pitfalls.
3.1. Readers, Writers, Tool Developers
Network management data modeling languages are used by engineers to
develop and specify data models for (sub-components of) managed
systems. The data models are later read by implementers with the
goal to create interoperable implementations as well as network
operators while managing networks and services. A data modeling
language is therefore used by both, data model writers and data model
readers. In most cases, the number of readers of a data model is
much larger than the number of writers. Hence, when taking design
decisions, it may be necessary to consider to what extend the design
choice simplifies the task of data model readers and the task of data
model writers. All other things being equal, preference should be
given to solutions that simplify the task of the human reader.
The third group to be considered in the data modeling language design
process are tool developers. In general, language features should be
simple and straight forward to implement in order to achieve wide
spread tool support. However, when considering alternatives, it is
important to realize that in most cases the number of data model
writers is much larger than the number of tool developers.
As a consequence, preference should be give to solutions that
simplify the task of readers. Reduction of the efforts required by
writers is of secondary importance and the reduction of the efforts
required by tool developers is of least important preference. While
this perhaps seems obvious, it is at times difficult to take
according design decisions if the group working on a new data
modeling language is dominated by say tool developers or data model
writers.
Schoenwaelder Expires December 17, 2007 [Page 4]
Internet-Draft Lessons Learned from the SMIng Project June 2007
3.2. Data vs. Command vs. Object vs. Document
There are different approaches to model network management interfaces
and to design supporting protocols:
o The data-oriented approach models all management aspects through
data objects that can be read, written, created, and destroyed.
The SNMP technology is an extreme example of this approach where
even create and destroy operations are handled as side effects of
write operations. While the data-oriented approach is
conceptually simple, it is usually difficult to use for modeling
complex operations, for example due to atomicity issues.
o The command-oriented approach does not model state objects but
instead treats the state maintained by a managed element as a
black box and provides specific commands to cause state changes or
to retrieve/write some selected state information. The command-
oriented approach is widely used by command line interfaces.
o The object-oriented approach can be seen as a combination of the
data-oriented and the command-oriented approach where commands are
bound to data objects and the black-box state is limited to the
internal state of objects. The object-oriented approach is being
used for example by the Common Information Model.
o The document-oriented approach represents the state of a device
and its configuration as a structured document. As a consequence,
primitive operations are the retrieval of (parts of) a document
and the application of changes to move from one document version
to the next. Since a configuration is treated as a complete
document, it becomes more natural to support course grained atomic
operations. The document-oriented approach is used by NETCONF for
dealing with configurations.
Note that the choice of the approach has impact on the features
required by a data modeling language. Trying to be protocol neutral
may imply to support multiple of these approaches, which clearly
increases the complexity of the language design. A data modeling
language aiming at supporting a document-oriented view for
configuration, a data-oriented view for monitoring, and a command-
oriented view for ad-hoc operations on a device must be able to link
say a command to the data objects that may be triggered to achieve
the behaviour defined for that command. This quickly becomes
difficult and it might be better to avoid mixing different approaches
for the same functionality.
3.3. Naming Independence
One of the goals of the SMIng project was to provide a data
definition language where the data definitions are not tied to a
specific management protocol. Since protocols typically use
Schoenwaelder Expires December 17, 2007 [Page 5]
Internet-Draft Lessons Learned from the SMIng Project June 2007
different approaches to name instances, it is required to support
multiple instance naming systems. While this may sound easy to do,
it turns out that trying to achieve naming independence has many
implications:
o There is in general a need to model relationships. A common
approach is to refer to other definitions using names as pointers.
Being naming system independent, such a pointer needs to be
abstracted and the protocol mappings then need to explain how the
pointer works for a given protocol. This becomes quickly
difficult since management protocols tend to introduce different
specific constraints.
o Many relationships between data model components are typically
explained in description clauses. Being protocol neutral,
description clauses have to be worded carefully so that they make
sense with the various protocol mappings. This significantly
increases the efforts to write data models and to review them.
o When thinking about relationships, it is crucial to think about
fate sharing of data objects, namely does the removal of an object
imply the removal of other objects? Being naming system
independent causes some additional challenges here since some fate
sharing properties may be implicit in the naming system used in
once protocol mapping while they must be explicit in other
protocol mappings.
3.4. Errors and Exceptions
Well written data definitions include clear definitions how
implementations should react in various error situations or during
run-time exceptions. It is often a mistake if a data model writer
assumes that implementers will choose a particular error code in a
given error situation. For interoperability reasons, it is far
better to spell out the precise behaviour in error situations and
run-time exceptions. Unfortunately, it becomes quite difficult to be
precise about errors and exceptions in a protocol and naming system
neutral data modeling framework such as SMIng since all error and
exception reporting mechanisms of the underlying protocols need to be
abstracted since otherwise description clauses become protocol
specific. Furthermore, due to the varying features of the underlying
protocols, some errors and exceptions may only exist in some of the
mappings and must therefore be dealt with when the generic data model
is mapped to a concrete protocol.
3.5. Configuration Storage Models
Management protocols use different mechanisms to deal with
persistence of configuration state. The SNMP protocol uses
StorageType [RFC2579] columns in conceptual tables to indicate which
Schoenwaelder Expires December 17, 2007 [Page 6]
Internet-Draft Lessons Learned from the SMIng Project June 2007
rows are persistent. The COPS-PR [RFC3084]protocol binds provisioned
information to the existence of the COPS-PR session which provisioned
policy information. The NETCONF protocol [RFC4741] has a concept of
different data stores (running, startup, candidate) that imply
whether configuration state is persistent or not.
Given these fundamentally different approaches, it is difficult to
write protocol neutral configuration data models that work equally
well with the various management protocols. While it is possible to
introduce say StorageType columns in protocol mappings, it is
sometimes awkward to define the semantics of data objects if the
persistence model is left to the protocol mappings.
3.6. Selection of Language Features
A data modeling language should be based on a small orthogonal set of
language features. In addition, the language should be minimalistic.
During the design of the language, it is useful to repeatedly ask the
question why certain features are needed or whether there are any
language rules without a strong justification. The general principle
stated by Antoine de Saint-Exupery holds:
Perfection [in design] is achieved, not when there is nothing more
to add, but when there is nothing left to take away.
Language features should be selected by considering what tools can do
with the information formalized by a language feature. In the
general case, language features should raise automation and improve
clarity. Features without a clear justification that they actually
reduce implementation efforts or significantly improve the clarity of
a data model specification should not be accepted.
3.7. Syntax versus Semantics
During the design of a data modeling language, it is most important
to focus on a clear semantic of all language constructs. Syntactical
issues are of secondary importance. While this may sound obvious, it
is important especially in open concensus driven processes to keep
contributors reminded that semantics of language constructs come
first while the syntactic aspects are of secondary importance.
3.8. Reuse of Definitions
A good data modeling language should allow data model writers to
produce reusable definitions. However, not all definitions are
equally easy to reuse. Experience has shown that data types derived
from a small set of base types are usually very easy to reuse;
structured data types are often getting slightly more complex to
Schoenwaelder Expires December 17, 2007 [Page 7]
Internet-Draft Lessons Learned from the SMIng Project June 2007
reuse. Reusing collections of structured data types with non-trivial
interrelationships are often difficult to reuse. The low hanging
fruits are therefor the simple type definitions.
Sub-typing is frequently used to add specific semantics to a more
generic type and to constrain the set of possible values. While
constraining values is often a good thing, it should be noted that
too strict constraints can turn out to be painful in the future.
There are many examples where data types proved too constrained to
represent protocols that followed a natural evolution.
3.9. Versioning and Meta Information
Changes to data models are inevitable. Even the most careful design
and review process will not be able to predict how technologies
evolve in the future. It is therefore crucial that the data modeling
language supports a suitable versioning model and that it established
clear rules which changes are possible without having to change a
version number or (module) name. Supporting complex collections of
meta data (e.g., a granular revision log) as part of the data
modeling language may be less of an issue since such information can
often be maintained in revision control systems and automatically
places into comments where needed.
3.10. Extensibility
It has proven to be useful to support language extensibility features
that avoid backward compatibility problems with existing parsers when
new language features are introduced. If a data modeling language
supports language extensions, it should be possible to gradually
introduction of new language features.
Supporting language extensions essentially requires to be very
consistent with the syntactic structure so that implementations can
skip over new language constructs. In addition, it is necessary to
precisely identify language extensions so that implementations
supporting extensions know where to apply the appropriate additional
processing.
3.11. Language Independence
Once way to define a new data modeling language is to start from an
existing language and to extend it and/or subset it as needed.
However, there are some dangers associated with this approach. Since
languages that are used are not static, it is important to deal with
conflicts that can arise if the base language is revised such that
the extensions or subsets do not work anymore. This issue becomes
particularly important if different organizations have change control
Schoenwaelder Expires December 17, 2007 [Page 8]
Internet-Draft Lessons Learned from the SMIng Project June 2007
since this makes coordination relatively complex. Things get even
worse if the data modeling extension constitutes not a significant
usage of the core language.
The alternative is to provide a complete and self-contained
definition for the data modeling language. This protects against
change control issues and also makes it simpler for implementors or
data model writers to find all language rules in one place.
As a general rule, there is hardly a reason to import language
definitions. If you like to import something from a source that is
likely to change, do not import. If, however, the source is assumed
to be stable, then you can just import by value and explain the
relationship to the original source.
3.12. Compliance and Conformance
Any successful data modeling language will at some point have to deal
with compliance and conformance definitions. However, getting
language features to express compliance and conformance statements
right is non-trivial. Very fine grained mechanisms to define
implementation requirements for different usage scenarios can lead to
very detailed but also very complex compliance and conformance
definitions that are difficult to understand and review.
It is therefore important to make a good tradeoff between the
required expressiveness and the pragmatic usage of compliance and
conformance definitions. The general considerations for language
features discussed Section 3.6 particularly apply here. One approach
is to first gain experience with the usage of a new data modeling
language and to support compliance and conformance definitions
through language extensions.
4. Conclusions
This memo documents some of the insights gained in the SMIng project
in order to guide future data modeling language designers by making
them aware of some of the problems encountered in the SMIng project
and some of the design guidelines that have been developed through
the project.
5. IANA Considerations
This document has no IANA actions.
Schoenwaelder Expires December 17, 2007 [Page 9]
Internet-Draft Lessons Learned from the SMIng Project June 2007
6. Security Considerations
This document discusses lessons learned during the design of a
network management data modeling language and they have no impact on
the security of the Internet.
7. Acknowledgements
Many people contributed directly or indirectly to this documented
through their participation in the SMIng project. Frank Strauss
provided valuable feedback on earlier versions of this document.
8. Informative References
[RFC2579] McCloghrie, K., Perkins, D., and J. Schoenwaelder,
"Textual Conventions for SMIv2", RFC 2579, April 1999.
[RFC3084] Chan, K., Seligson, J., Durham, D., Gai, S., McCloghrie,
K., Herzog, S., Reichmeyer, F., Yavatkar, R., and A.
Smith, "COPS Usage for Policy Provisioning (COPS-PR)",
RFC 3084, March 2001.
[RFC3780] Strauss, F. and J. Schoenwaelder, "SMIng - Next Generation
Structure of Management Information", RFC 3780, May 2004.
[RFC3781] Strauss, F. and J. Schoenwaelder, "Next Generation
Structure of Management Information (SMIng) Mappings to
the Simple Network Management Protocol (SNMP)", RFC 3781,
May 2004.
[RFC3444] Pras, A. and J. Schoenwaelder, "On the Difference between
Information Models and Data Models", RFC 3444,
January 2003.
[RFC4741] Enns, R., "NETCONF Configuration Protocol", RFC 4741,
December 2006.
Schoenwaelder Expires December 17, 2007 [Page 10]
Internet-Draft Lessons Learned from the SMIng Project June 2007
Author's Address
Juergen Schoenwaelder
Jacobs University Bremen
Campus Ring 1
28725 Bremen
Germany
Phone: +49 421 200-3587
Email: j.schoenwaelder@jacobs-university.de
Schoenwaelder Expires December 17, 2007 [Page 11]
Internet-Draft Lessons Learned from the SMIng Project June 2007
Full Copyright Statement
Copyright (C) The IETF Trust (2007).
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, THE IETF TRUST 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.
Intellectual Property
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-ipr@ietf.org.
Acknowledgment
Funding for the RFC Editor function is provided by the IETF
Administrative Support Activity (IASA).
Schoenwaelder Expires December 17, 2007 [Page 12]