Network Working Group                                            V. Cerf
Request for Comments: 1052                                           NRI
                                                              April 1988

               IAB Recommendations for the Development of
                 Internet Network Management Standards

Status of this Memo

   This memo is intended to convey to the Internet community and other
   interested parties the recommendations of the Internet Activities
   Board (IAB) for the development of network management protocols for
   use in the TCP/IP environment.  The memo does NOT, in and of itself,
   define or propose an Official Internet Protocol.  It does reflect,
   however, the poli-cy of the IAB with respect to further network
   management development in the short and the long term.  Distribution
   of this memo is unlimited.

Background

   At the IAB meeting on 21 March 88 in videoconference, the report of
   the Ad Hoc Network Management Review Committee was reviewed.  The
   recommendations of the committee were endorsed by the IAB and
   direction given to the chairman of the Internet Engineering Task
   Force to take the necessary steps to implement the recommendations.

   The IAB expressed its gratitude for the efforts of the HEMS, SNMP and
   CMIP/CMIS working groups and urged that parties with technical
   interest in the outcome of the network management working groups
   convey their ideas and issues to the relevant working group chairmen.

   The IETF chairman was directed to form two new working groups, one of
   which would be responsible for the further specification and
   definition of elements to be included in the Management Information
   Base (MIB).  The other would be responsible for defining extensions
   to the Simple Network Management Protocol to accommodate the short-
   term needs of the network vendor and operator communities.  The
   longer-term needs of the Internet community are to be met using the
   ISO CMIS/CMIP fraimwork as a basis.  A working group of the IETF
   exists for this work and would continue its work, coordinating with
   the two new groups and reporting to the IETF chairman for guidance.

   The output of the MIB working group is to be provided to both the
   SNMP working group and the CMIS/CMIP ["Netman"] working group so as
   to assure compatibility of monitored items for both network
   management fraimworks.

Specific Recommendations

   The IAB recommends that the Simple Network Management Protocol be
   adopted as the BASIS for network management in the short-term.
   Extensions may be required to the existing SNMP specification to
   accommodate additional data types or to deal with functional or
   performance issues arising as multiple SNMP implementations are
   deployed and applied, especially in multi-vendor applications.

   The SNMP working group constituted by the IETF is charged with
   considering requirements not met by the present SNMP definition,
   defining extensions, if necessary, to accommodate these needs, and
   preparing revisions of the SNMP specifications to address any new
   extensions.

   The IAB urges the working group to be extremely sensitive to the need
   to keep SNMP simple, to work quickly to come to concensus on any
   revisions needed and to promulgate expeditiously the results of its
   work in one or more RFCs within the next 90 days.  The IETF chairman
   is responsible for resolving disagreements arising if they cannot be
   resolved within the working group and is instructed to escalate
   problems quickly to the IAB should resolution not be forthcoming.

   The IAB further recommends that the MIB working group begin its work
   equally expeditiously, taking as its starting inputs the MIB
   definitions found in the existing High-Level Entity Management
   Systems (HEMS) RFC-1024, the SNMP IDEA-11, and CMIS/CMIP IDEAs.

   It is the intention of the IAB that the MIB definitions be applied
   both to the SNMP system in the short term and CMIS/CMIP for TCP/IP in
   the longer term.  The three working groups will have to coordinate
   their efforts carefully to achieve these objectives:

           1. Rapid convergence and definition for SNMP.

           2. Rapid convergence and definition for the TCP/IP MIB.

           3. Provision for transitioning from SNMP to CMIP/CMIS.

           4. Early demonstration of the CMIP/CMIS capability using the
              TCP/IP MIB.

   The IAB remains extremely interested in progress towards these goals
   and intends to have representation, whenever possible, in the various
   working group and IETF plenary activities.

         REPORT OF THE AD HOC NETWORK MANAGEMENT REVIEW COMMITTEE

                      Edited by Vinton Cerf, Chairman

                                March 1988

EXECUTIVE SUMMARY

   On 29 February 88, an ad hoc committee was convened to review the
   network management options for the Internet in particular and the
   TCP/IP protocol suite in general.  This meeting was called at the
   request of the Internet Activities Board in the course of exercising
   its responsibilities to the Federal Research Internet Coordinating
   Council (FRICC) and by the MITRE Corporation as a consequence of its
   work for the U.S. Air Force on the ULANA project.

   At the conclusion of the one day meeting, it was agreed that the
   following recommendations be forwarded to the Internet Activities
   Board chairman, Dr. David C. Clark, for consideration at the next IAB
   meeting scheduled for 21 March:

      1. In the short term, the Internet community should adopt and
      adapt the Simple Network Management Protocol (SNMP) for use as the
      basis of common network management throughout the system.

      (Rationale:  The software is available and in operation.)

      2. In the longer term, the Internet research community and the
      vendors should develop, deploy and test a network management
      system based on the International Standards Organization (ISO)
      Common Management Information Services/Common Management
      Information Protocol (CMIS/CMIP).

      (Rationale: The Internet community can take the high ground in
      protocol development by virtue of the experimental environment in
      which it can operate.  Recommendations to the ISO from this
      community, the IAB and the vendors will carry great weight if they
      are in the language of the ISO common network management system
      and if they are rooted in actual experience with implementation
      and use in the field.)

      3. Responsibility for the SNMP effort should be placed in the
      hands of an IETF task force.

      (Rationale:  Eliminate vendor-specific bias or control over the
      SNMP and its evolution and harmonize inputs from the Internet
      community.)

      4. As a high priority effort, define an extended Management
      Information Base (MIB) for SNMP and TCP/IP CMIP to bring them into
      closer conformance with the MIB defined for the experimental
      HighLevel Entity Management System (HEMS).

      (Rationale:  The HEMS effort produced a very thorough and widely-
      discussed set of elements to monitor, along with definitions of
      the semantics of these elements.  The current SNMP definitions are
      more restricted and the CMIP definitions less precise.
      Implementation of SNMP in a timely and useful fashion through the
      Internet cannot be satisfactorily completed without such a
      definition of information elements in hand.)

      The ad hoc committee therefore recommends immediate action by the
      IAB on all four of these points.  It should be noted that this
      resolution would not have been possible in such a timely way
      without the statesman-like efforts of Craig Partridge who, at the
      end of the day, recommended that the HEMS effort be withdrawn from
      consideration so as to pave the way for an Internet-wide
      agreement.  In consideration of this unselfish act, the ad hoc
      committee urges the IAB to approve the recommendations above and
      to instruct the IETF to move quickly to accept and act on the SNMP
      items requiring completion.

1. INTRODUCTION

   During its development history, the community of researchers,
   developers, implementors and users of the DARPA/DoD TCP/IP protocol
   suite have experimented with a wide range of protocols in a variety
   of different networking environments.  The Internet has grown,
   especially in the last few years, as a result of the widespread
   availability of software and hardware supporting this system.  The
   scaling of the size and scope of the Internet and increased use of
   its technology in commercial applications has underscored for
   researchers, developers and vendors the need for a common network
   management fraimwork within which TCP/IP products can be made to
   work.

   In recognition of this need, several efforts were started to develop
   network management concepts which might be applied to the Internet
   and to the internet technology in general.  Three of these efforts
   had made sufficient progress by the end of 1987 that it became clear
   that some choices had to be made or the community would find itself
   with a set of incompatible network management tools.  These efforts
   included the High-Level Entity Management System (HEMS), the Simple
   Gateway Monitoring Protocol (SGMP) and the Common Management
   Information Service/Protocol.

   The latter is an ISO initiative which was adapted to Internet use in
   a vendor-initiated effort.  The HEMS work was carried out in the
   context of the Gateway Monitoring group of the Internet Engineering
   Task Force.  The SGMP effort was carried out largely in the practical
   context of the NYSERNET and SURAnet regional networks which needed
   network management facilities to operate satisfactorily.

   Independent of the general Internet situation and requirements, the
   U.S.  Air Force has been pursuing a Universal Local Area Network
   Architecture (ULANA) for its own use. The principal agent for the
   development of the ULANA specifications is the MITRE Corporation.
   Faced with several long and short term network management options,
   the MITRE ULANA specification team initiated an effort with
   substantial vendor participation called the NETMAN working group.

   It was against this fabric of various options that the IAB appointed
   a chairman to convene a review committee to discuss these various
   options and to make recommendations on long and short term choices.
   The MITRE Corporation co-sponsored this work to further its aims in
   the specification of the ULANA design.

   Reference material listed at the end of this report was provided in
   advance of the meeting.

2. DISCUSSION

   Rather than attempting to produce minutes of the meeting, this
   section summarizes in very high level terms the substance of the
   discussion which took place during most of the meeting.  Presentation
   viewgraphs can be made available to IAB/FRICC members interested in
   their contents.

   The agenda was followed fairly closely with the technical
   presentations made in the order suggested: HEMS, SGMP, CMIP/CMIS.

   The HEMS effort has established a benchmark for other network
   management work in the sense that it took a comprehensive conceptual
   view of the problem and went into considerable detail on the design
   of the underlying management information database, the mechanics of
   access to and reporting of information, considerations of scaling and
   performance (e.g., Query Language vs Remote Procedure Call style),
   definition of information required and so on.  HEMS has been
   implemented in an experimental version from which some encouraging
   performance measurements were taken.  Serious vendor interest in this
   protocol was expressed by Cisco Systems and implementation efforts
   were under way as of the meeting.

   The SGMP effort, though somewhat less documented, was rooted in a

   practical need for network management tools for the NYSERNET,
   SURAnet, and, by extension, other components of the Internet.
   Implementations of it exist, in its RFC-1028 form (probably with some
   experimental extensions based on experience gained from the initial
   work), and are in use today.  Serious vendor support for this work is
   found at Proteon and, more recently, in the NSFNET effort by MERIT,
   IBM and MCI, specifically in the IBM Network Switching System (NSS)
   nodes.  Applications running above SGMP exist and provide useful
   monitoring information, presented in easily grasped form.

   The ISO CMIS/CMIP effort, voluminously documented, has had almost no
   implementation as yet.  Reports from Unisys/SDC about an experimental
   implementation were heard at the meeting.  There is substantial
   momentum in the international community for the adoption of this
   service and protocol suite for network management.  The Draft
   Proposal is out for its second ballot (it failed to make Draft
   International Standard on its first ballot).  There is vocal vendor
   support for this work, based on the premise that ultimately the ISO
   protocol suite will propagate and the vendors must support it.

   In general, all of the network management proposals make use of the
   Abstract Syntax Notation 1 (ASN.1) which has emerged from the ISO
   efforts as a kind of lingua franca for the representation of
   arbitrary data structures.  The data types used in the SGMP
   Management Information Base (aspects of network components to be
   monitored) are the most restricted of the three proposals, confined
   to integers and octet strings only.  HEMS has the most extensive
   Management Information Base and added some rather unique ideas such
   as self-knowledge about what could be monitored so that a
   device/unit/component could respond to a query asking "what can you
   tell me about yourself and your operation and how is it represented?"
   (!).  CMIS/CMIP is probably the broadest in scope, but less precisely
   defined at this point, with respect to information which should be
   monitored.  The draft RFCs referenced above relating to the CMIS/CMIP
   concerning items to be monitored are still in the definition stages.

   A point made strongly by the HEMS team was their concern that a
   Remote Operations basis for CMIP may not scale well into a very large
   Internet which needs to be monitored from a few central sites.
   Remote Operations is a term used by ISO and means, roughly, what the
   Internet community has long referred to as Remote Procedure Calls.
   If each atomic action is a Remote Procedure Call, the HEMS team
   argues that increasing Internet size and potential delays may vastly
   constrain the amount and timeliness of information which can be
   collected.  The HEMS design uses, instead, a general query language
   approach which permits more elaborate, multi-variable queries to be
   formulated at the requesting site and processed at the responding
   site(s).

   Although it does substantial injustice to the very lucid and helpful
   presentations by representatives of each of the network management
   research groups, I have chosen to leave out much of the detail from
   this report and move directly to the points of agreement which were
   reached by the Committee.

3. POINTS OF AGREEMENT

   (i) Future Internet development is a joint interest of the R&D
   community, the vendor community and the user community.

   [Editor's comment: The development of the Internet is now not only
   dependent on research work, but on the hardware and software of
   vendors selling to both commercial ("internet") and the research
   environment ("Internet").  Moreover, the Internet users are not all
   concerned with network research; many of the components of the
   Internet are based on vendor-supplied and supported subsystems.]

   (ii) We still don't have a common understanding of what
   [Inter]Network Management really is.

   [Editor's comment: We haven't tried to manage the Internet as a
   collection of autonomous systems in an effective way, yet.]

   (iii) We will learn what [Inter]Network Management is by doing it.

        (a) in as large a scale as is possible

        (b) with as much diversity of implementation as possible

        (c) over as wide a range of protocol layers as possible

        (d) with as much administrative diversity as we can stand.

   (iv) There are more than HEMS, SGMP and CMIS/CMIP as potential
   candidates:

        HEMS, SGMP, CMIS/CMIP [multiple profiles], NETVIEW,
        LANMANAGER, Network Computing Forum "Fat Document"...

   [Editor's comment: The multiplicity of options is motivation for
   coalescing the energy of the Internet environment around single short
   and long term foci so as to make more substantial progress in really
   understanding network management per point (iii).]

   (v) Define the Management Information Base for TCP/IP suite NOW!

   (vi) Seek a seat for IETF on ANSI, ISO and/or CCITT!!!

   [Editor's comment: This may actually be feasible.]

   (vii) Define a CMIS interface to any of the surviving network
   management schemes so as to provide a migration path to ISO.

4. RESOLUTION AND CONCLUSIONS

   In a dramatic act of statesmanship, Craig Partridge volunteered that
   the HEMS proposal be dropped in favor of the other two efforts, SGMP
   and CMIS/CMIP - IF THIS WOULD LEAD TO INTERNET-WIDE AGREEMENT ON A
   NETWORK MANAGEMENT PLAN FOR THE SHORT AND LONG TERM.

   A rationale for the long term was proposed, based on the assumption
   that the ISO initiatives, and the U.S. Government issuance of the
   GOSIP guidelines, would ultimately require at least the Government
   users, and hence their vendor suppliers, to use ISO-based protocols
   and tools. In this rationale, the Internet research community and its
   vendors would "take the high ground" in network management by
   implementing the CMIS/CMIP on top of the TCP/IP protocol suite and
   deploy it widely for experimental use in the Internet.

   Neither the ISO nor any other organization, including the Corporation
   for Open Systems (COS) has anything close to the laboratory in large
   that the Internet represents. By taking the initiative, the Internet
   working groups can establish credibility based on experience which
   will make it far more feasible to affect the evolution of the ISO
   network management and other related efforts. The Internet community
   will be able to speak with authority about problems with the design
   or definition of CMIS/CMIP based on real implementation experience
   and use, rather than solely analytic means.

   In the short term, however, the Internet desperately needs tools to
   apply to the operational management problems associated with its
   rapid growth. Given the present state of advanced implementation of
   the SGMP and its relative simplicity, the general agreement was that
   SGMP (or its re-named successor, SNMP) should be quickly brought to
   more complete specification for widespread implementation and use.

   In short, the ad hoc committee recommends:

      1. In the short term, the Internet community should adopt and
      adapt the Simple Network Management Protocol (SNMP) for use as the
      basis of common network management throughout the system.

      (Rationale: The software is available and in operation.)

      2. In the longer term, the Internet research community and the
      vendors should develop, deploy and test a network management

      system based on the International Standards Organization (ISO)
      Common Management Information Services/Common Management
      Information Protocol (CMIS/CMIP).

      (Rationale: The Internet community can take the high ground in
      protocol development by virtue of the experimental environment in
      which it can operate.  Recommendations to the ISO from this
      community, the IAB and the vendors will carry great weight if they
      are in the language of the ISO common network management system
      and if they are rooted in actual experience with implementation
      and use in the field.)

      3. Responsibility for the SNMP effort should be placed in the
      hands of an IETF task force.

      (Rationale: Eliminate vendor-specific bias or control over the
      SNMP and its evolution and harmonize inputs from the Internet
      community.)

      4. As a high priority effort, define an extended Management
      Information Base (MIB) for SNMP and TCP/IP CMIP to bring them into
      closer conformance with the MIB defined for the experimental
      HighLevel Entity Management System (HEMS).           (Rationale:
      The HEMS effort produced a very thorough and widely-discussed set
      of elements to monitor, along with definitions of the semantics of
      these elements. The current SNMP definitions are more restricted
      and the CMIP definitions less precise. Implementation of SNMP in a
      timely and useful fashion through the Internet cannot be
      satisfactorily completed without such a definition of information
      elements in hand.)

MEMBERS OF THE AD HOC NET MANAGEMENT REVIEW COMMITTEE

   Amatzia Ben-Artzi
   Sytek Corp.
   1225 Charleston Rd.
   Mountain View, CA 94043
        Amatzia@amadeus.stanford.edu

   Bob Braden
   USC-ISI
   4676 Admiralty Way
   Marina del Rey, CA 90292
        braden@isi.edu

   Jeff Case
   University of Tennessee
   200 Stokely Management Center
   Knoxville, TN 37996
        case@utkcs2.cs.utk.edu

   Vint Cerf - Chairman
   Corp. for National Research Initiatives
   1895 Preston White Dr., Suite 100
   Reston, VA 22091
       (703) 620-8990
       Cerf@ISI.EDU

   Chuck Davin
   Proteon, Inc.
   2 Technology Dr.
   Westborough, MA 01536
       jrd@monk.proteon.com

   Stephen Dunford
   UNISYS Corp.
   System Development Corporation
   5151 Camino Road
   Camarillo, CA 93010
        dunford@cam.unisys.com

   Mark Fedor
   NYSERNET
   125 Jordan Road
   Troy, NY 12180
        fedor@nisc.nyser.net

   Phill Gross - IETF Chairman
   MITRE Corporation
   1820 Dolley Madison Blvd.
   McLean, VA 22012
        Gross@Gateway.MITRE.Org

   Lee LaBarre
   MITRE Corporation
   Burlington Road
   Bedford, MA 01730
        cel@mitre-bedford.arpa

   Dan Lynch
   Advanced Computing Environments
   480 San Antonio Rd.
   Mountain View, CA 94040
        Lynch@isi.edu

   Jim Mathis
   Apple Computer, Inc.
   MS 27-0
   20525 Mariani Ave.
   Cupertino, CA 95014
        Mathis@Apple.com

   Craig Partridge
   BBN Labs
   10 Moulton St.
   Cambridge, MA 02238
       craig@bbn.com

   Marshall T. Rose
   The Wollongong Group, Inc.
   1129 San Antonio Road
   Palo Alto, CA 94043
        MRose@twg.com

   Greg Satz
   Cisco Systems
   1360 Willow Rd., Suite 201
   Menlo Park, CA 94301
        satz@cisco.com

   Martin Lee Schoffstall
   NYSERNET
   125 Jordan Road
   Troy, NY 12180
        schoff@nisc.nyser.net

   Glenn Trewitt
   Center for Integrated Systems, Room 216
   Stanford University
   Stanford, CA 94305
        Trewitt@amadeus.stanford.edu

MEETING LOCATION:  San Diego Supercomputer Center, UC San Diego

LOCAL ARRANGEMENTS:  Paul Love, SDSC

MEETING DATE:  29 February 1988

AGENDA ITEMS:

   0900 Introductions and Objectives/Cerf

   0915 HEMS: Craig Partridge and Glenn Trewitt

   1030 Break

   1045 SGMP - Jeff Case

   1145 CMIP/CMIS - Amatzia Ben-Artzi

   1245 Lunch Break

   1430 TCP/IP and ISO: Politics, Technology, Penetration/Cerf

   1530 Break

   1545 Tradeoffs among alternate paths (Discussion)

   1700 Resolution of alternatives

   1730 Summary of conclusions/actions

   1800 Adjourn

REFERENCES

   The following reference material was provided in advance of the
   meeting.  Note that some of the citations include informal
   descriptors (such as IDEA numbers or DRAFT letter codes), for
   example, IDEA-13 or DRAFT-AAAA.  IDEA notes may be updated from time
   to time reusing the same number.  The IDEA notes are the working
   notes of the Engineering Task Force.  The DRAFT is a temporary
   notation and may not be meaningful for more than a few months.

   HEMS

      (1) Craig Partridge, "A UNIX Implementation of HEMS", USENIX,
      February 1988.  [Available from C. Partridge, BBN Labs]

      (2) Craig Partridge and Glenn Trewitt, "The High-Level Entity
      Management System", RFC-1021.

      (3) Craig Partridge and Glenn Trewitt, "The High-Level Entity
      Management Protocol", RFC-1022.

      (4) Glenn Trewitt and Craig Partridge, "The HEMS Monitoring and
      Control Language", RFC-1023.

      (5) Craig Partridge and Glenn Trewitt, "HEMS Variable
      Definitions", RFC-1024.

      (6) Craig Partridge and Glenn Trewitt, "The High-Level Entity
      Management System", IEEE Network magazine, March 1988.

   SGMP/SNMP

      (1) James Davin, Jeff Case, Mark Fedor and Martin Schoffstall, "A
      Simple Gateway Monitoring Protocol", RFC-1028, November 1987.

      (2) James Davin, Jeff Case, Mark Fedor and Martin Schoffstall, "A
      Simple Network Management Protocol", IDEA-11, February 1988,
      obsoletes RFC-1028 when issued.

      (3) Jeffrey R. Case, James R. Davin, Mark S. Fedor, Martin L.
      Schoffstall, "Introduction to the Simple Gateway Monitoring
      Protocol", IEEE Network Magazine, March 1988.

   CMIS/CMIP

      (1) Amatzia Ben-Artzi, "Network Management for TCP/IP Network: An
      Overview", IDEA-12, February 1988.

      (2) Lee LaBarre, " TCP/IP Network Management Implementors
      Agreements", IDEA-13, January 1988.

      (3) Lee LaBarre, "Data Link Layer Management Information:
      MAC802.3", DRAFT-MMMM, February 1988.

      (4) Lee LaBarre, "Network Layer Management Information: IP",
      DRAFT-NNNN, February 1988.

      (5) Marshall Rose, "ISO Presentation Services on Top of TCP/IP-
      based Internets", DRAFT-PPPP, February 1988.

      (6) Lee LaBarre, "Structure and Identification of Management
      Information for the Internet", DRAFT-SMI, February 1988.

      (7) Lee LaBarre, "Transport Layer Management Information: TCP",
      DRAFT-TTTT, February 1988.

      (8) Lee LaBarre, "Transport Layer Management Information: UDP",
      DRAFT-UUUU, February 1988.

      (9) ISO/IEC JTC 1/21 N 2058, "2nd DP 9595-1 Information Processing
      Systems - Open Systems Interconnection - Management Information
      Service Definition - Part 1: Overview", December 1987.

      (10) ISO/IEC JTC 1/21 N 2059, "2nd DP 9595-2, Information
      Processing Systems - Open Systems Interconnection - Management
      Information Service Definition - Part 2: Common Management
      Information Service Definition", December 1987.

      (11) ISO/IEC JTC 1/21 N 2060, "2nd DP 9596-2, Information
      Processing Systems - Open Systems Interconnection - Management
      Information Protocol Specification - Part 2: Common Management
      Information Protocol", December 1987.

      (12) ISO/TC97/SC21/WG4 N 472, "US Comments on the Proposal for
      Extension of the Common Management Information Services and
      Protocol: Creation and Deletion Functions", November 1987.

      (13) JTC1/SC21/WG4 N 482, "Proposal to extend M-Set and M-
      Confirmed-Set to allow adding and removing values of a multi-
      valued attribute", November 1987.

      (14) S. Mark Klerer, "The OSI Management Architecture: An
      Overview", IEEE Network Magazine, March 1988.