Internet Engineering Task Force                   Integrated Services WG
INTERNET-DRAFT                              S. Shenker and J. Wroclawski
draft-ietf-intserv-charac-02.txt                      Xerox PARC/MIT LCS
                                                           October, 1996
                                                           Expires: 3/97


                General Characterization Parameters for
                  Integrated Service Network Elements


Status of this Memo


   This document is an Internet-Draft. 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.''

   To learn the current status of any Internet-Draft, please check the
   ``1id-abstracts.txt'' listing contained in the Internet-Drafts Shadow
   Directories on ftp.is.co.za (Africa), nic.nordu.net (Europe),
   munnari.oz.au (Pacific Rim), ds.internic.net (US East Coast), or
   ftp.isi.edu (US West Coast).

   This document is a product of the Integrated Services working group
   of the Internet Engineering Task Force. Comments are solicited and
   should be addressed to the working group's mailing list at int-
   serv@isi.edu and/or the author(s).

Abstract

      This memo defines a set of general control and characterization
      parameters for network elements supporting the IETF integrated
      services QoS control framework. General parameters are those with
      common, shared definitions across all QoS control services.

1. Introduction

   This memo defines the set of general control and characterization
   parameters used by network elements supporting the integrated
   services framework.  "General" means that the parameter has a common
   definition and shared meaning across all QoS control services.



Shenker/Wroclawski         Expires March, 1997                  [Page 1]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   Control parameters are used by applications to provide information to
   the network related to QoS control requests. An example is the
   traffic specification (TSpec) generated by application senders and
   receivers.

   Characterization parameters are used to discover or characterize the
   QoS management environment along the path of a packet flow requesting
   active end-to-end QoS control.  These characterizations may
   eventually be used by the application requesting QoS control, or by
   other network elements along the path. Examples include information
   about which QoS control services are available along a network path
   and estimates of the available path bandwidth.

   Individual QoS control service specifications may refer to these
   parameter definitions as well as defining additional parameters
   specific to the needs of that service.

   Parameters are assigned machine-oriented ID's using a method
   described in [RFCTEMPLATE] and summarized here.  These ID's may be
   used within protocol messages and management interfaces to describe
   the parameter values present. Each parameter ID is composed from two
   numbers, one identifying the service associated with the parameter
   (the <service_number>), and the other (the <parameter_number>)
   identifying the parameter itself. <parameter_number> values for the
   parameters defined in this document are assigned from the "general
   parameters" range (1 - 127). Numbers in this range indicate that the
   parameter definition is shared by all QoS control services.

      NOTE: Parameters numbered 128 - 254 have definitions specific to a
      particular QoS control service. In contrast to the general
      parameters described here, the parameter's meaning depends on the
      service_number portion of the parameter ID.

      Service number 1 is reserved for use as described in Section 2 of
      this note. Service numbers 2 through 254 will be allocated to
      individual QoS control services. Currently, Guaranteed service
      [RFCG] is allocated number 2, and Controlled-load service [RFCCL]
      is allocated number 5.

   In this note, the textual form

     <service_number, parameter_number>

   is used to write a service_number, parameter_number pair.  The range
   of possible of service_number and parameter_number values specified
   in [RFCTEMPLATE] allow the parameter ID to directly form the tail
   portion of a MIB object ID representing the parameter. This
   simplifies the task of making parameter values available to network



Shenker/Wroclawski         Expires March, 1997                  [Page 2]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   management applications.

   The definition of each parameter used to characterize a path through
   the network describes two types of values; local and composed.  Local
   values reflect conditions at a single network element.  Composed
   values reflect the running composition of local values along a path,
   as specified by a composition rule.  Each parameter definition
   specifies the composition rule for that parameter. The composition
   rule describes how to combine the arriving composed value and the
   local value, to give a new composed value which is passed to the next
   network element in the path. Note that the composition may proceed
   either downstream, toward the receiver(s), or upstream, toward the
   sender. Each parameter may give only one definition for the local
   value, but may give more than one definition for composition rules
   and composed values. This is because it may be useful to compose the
   same local value several times following different composition rules.

   Because characterization parameters are used to compute the
   properties of a specific path through the internetwork, all
   characterization parameter definitions are conceptually "per-next-
   hop", as opposed to "per interface" or "per network element".  In
   cases where the network element is (or is controlling) a shared media
   or large-cloud subnet, the element may need to provide different
   values for different next-hops.  In practice, it may be appropriate
   for vendors to choose and document a tolerance range, such that if
   all next-hop values are within the tolerance range only a single
   value need be stored and provided.

   Local and composed characterization parameter values have distinct
   ID's so that a network management entity can request the value of
   either a local or path-composed parameter at any point within the
   network.

   Each parameter definition includes a description of the minimal
   properties, such as range and precision, required of any wire
   representation of that parameter's values. Each definition also
   includes an XDR [RFCXDR] description of the parameter, describing an
   appropriate external (wire) data representation for the parameter's
   values. This dual definition is intended to encourage a common wire
   representation format whenever possible, while still allowing other
   representations when required by the specific circumstances (e.g.,
   ASN.1 within SNMP).

   The message formats specified in [RFCRSVPUSE] for use with the RSVP
   setup protocol use the XDR data representation parameters.

   All of the parameters described in this note are mandatory, in the
   sense that a network element claiming to support integrated service



Shenker/Wroclawski         Expires March, 1997                  [Page 3]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   must recognize arriving values in setup and management protocol
   messages, process them correctly, and export a reasonable value in
   response. For certain parameters, it is required that the network
   element compute and export an *accurate* local value. For other
   parameters, it is acceptable for the network element to indicate that
   it cannot compute and export an accurate local value. The definition
   of these parameters provides a reserved value which indicates
   "indeterminate" or "invalid". This value signals that an element
   cannot process the parameter accurately, and consequently that the
   result of the end-to-end composition is also questionable.

      NOTE (temporary): Previous versions of this and the RSVP use
      document used both the reserved-value approach and a separate
      INVALID flag to record this fact.  Now, the reserved-value
      approach is used exclusively. This is so that any protocol which
      retrieves a parameter value, including SNMP, can carry the invalid
      indication without needing a separate flag. The INVALID flag
      remains in the RSVP message format but is reserved for use only
      with a possible future service-composition scheme.

2. Default and Service-Specific Values for General Parameters

   General parameters have a common *definition* across all QoS control
   services. Frequently, the same *value* of a general parameter will be
   correct for all QoS control services supported at a network element.
   In this circumstance, there is no need to export a separate copy of
   the value for each QoS control service; instead the node can export
   one copy which applies to all supported services.

   A general parameter value which applies to all services supported at
   a network node is called a default or global value. For example, if
   all of the QoS control services provided at a node support the same
   maximum packet size, the node may export a single default value for
   the PATH_MTU parameter described in Section 3, rather than providing
   a separate copy of the value for each QoS control service. In the
   common case, this reduces both message size and processing overhead
   for the setup protocol.

   Occasionally an individual service needs to report a value differing
   from the default value for a particular general parameter. For
   example, if the implementation of Guaranteed Service [RFCG] at a
   router is restricted by scheduler or hardware considerations to a
   maximum packet size smaller than supported by the router's best-
   effort forwarding path, the implementation may wish to export a
   "service-specific" value of the PATH_MTU parameter so that
   applications using the Guaranteed service will function correctly.
   In the example above, the router might supply a value of 1500 for the
   default PATH_MTU parameter, and a value of 250 for the PATH_MTU



Shenker/Wroclawski         Expires March, 1997                  [Page 4]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   parameter applying to guaranteed service. In this case, the setup
   protocol providing path characterization carries (and delivers to the
   application) both a value for Guaranteed service and a value for
   other services.

   The distinction between default and service-specific parameter values
   makes no sense for non-general parameters (those defined by a
   specific QoS control service, rather than this note), because both
   the definition and value of the parameter are always specific to the
   particular service.

   The distinction between default and service-specific values for
   general parameters is reflected in the parameter ID name space.  This
   allows network nodes, setup protocols, and network management tools
   to distinguish default from service-specific values, and to determine
   which service a service-specific parameter value is associated with.

   Service number 1 is used to indicate the default value. A parameter
   value identified by the ID:

     <1, parameter_number>

   is a default value, which applies to all services unless it is
   overridden by a service-specific value for the same parameter.

   A parameter value identified by the ID:

     <service_number, parameter_number>

   where service_number is not equal to 1, is a service-specific value.
   It applies only to the service identified by service_number.

   These service-specific values are also called override values.  When
   both service-specific and default values are present for a parameter,
   the service-specific value overrides the default value (for the
   service to which it applies). The rules for composing service-
   specific and global general parameters support this override
   capability.  The basic rule is to use the service-specific value if
   it exists, and otherwise the global value.

   A complete summary of the characterization parameter composition
   process is given below. In this summary, the "arriving value" is the
   incompletely composed parameter value arriving from a neighbor node.
   The "local value" is the (global or service-specific) value made
   available by the local node. The "result" is the newly composed value
   to be sent to the next node on the data path.

     1. Examine the <service_number, parameter_number> pair associated



Shenker/Wroclawski         Expires March, 1997                  [Page 5]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


     with the arriving value. This information is conveyed by the setup
     protocol together with the arriving value.

     2. If the arriving value is for a parameter specific to a single
     service (the parameter_number is larger than 128), compose the
     arriving value with the locally value exported by the specified
     service, and pass the result to the next hop.

     3. If the arriving value is a service-specific value for a general
     parameter (the parameter_number is 127 or less, and the
     service_number is other than 1), and the local implementation of
     that service also exports a service-specific value for the
     parameter, compose the service-specific arriving value and the
     service-specific local value of the parameter, and pass the result
     as a service-specific value to the next-hop node.

     4. If the arriving value is a service-specific value for a general
     parameter (the parameter_number is 127 or less, and the
     service_number is other than 1), and the local implementation of
     that service does *not* export a service-specific value, compose
     the service-specific arriving value with the global value for that
     parameter exported by the local node, and pass the result as a
     service-specific value to the next-hop node.

     5. If the arriving value is a global value for a general parameter
     (parameter_number is 127 or less, and the service_number is 1), and
     the local implementation of *any* service exports a service-
     specific value for that general parameter, compose the arriving
     (global) value with the service-specific value for that parameter
     exported by the local service, and pass the result as a service-
     specific value to the next-hop node. This will require adding a new
     data field to the message passed to the next hop, to hold the newly
     generated service-specific value. Repeat this process for each
     service that exports a service-specific value for the parameter.

     6. If the arriving value is a global value for a general parameter
     (the service_number is 1, and the parameter_number is 127 or less),
     compose the arriving (global) value with the global parameter value
     exported by the local node, and pass the result as a global
     (service 1) value to the next-hop node. This step is performed
     whether or not any service-specific values were generated and
     exported in step 5.

3. General Parameter Definitions







Shenker/Wroclawski         Expires March, 1997                  [Page 6]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


 3.1 NON-IS_HOP flag parameter

   This parameter provides information about the presence of network
   elements which do not implement QoS control services along the data
   path.

   The local value of the parameter is 1 if the network element does not
   implement the relevant QoS control service, or knows that there is a
   break in the chain of elements which implement the service.  The
   local parameter is 0 otherwise.  The local parameter is assigned
   parameter_number 1.

   The composition rule for this parameter is the OR function. A
   composed parameter value of 1 arriving at the endpoint of a path
   indicates that at least one point along the path does not offer the
   indicated QoS control service.  The parameter_number for the composed
   quantity is 2.

   The global NON_IS_HOP flag parameter thus has the ID <1,2>. If this
   flag is set, it indicates that one or more network elements along the
   application's data path does not support the integrated services
   framework at all. An example of such an element would be an IP router
   offering only best-effort packet delivery and not supporting any
   resource reservation requests.

   Obviously, a network element which does not support this
   specification will not know to set this flag.  The actual
   responsibility for determining that a network node does not support
   integrated services may fall to the network element, the setup
   protocol, or a manual configuration operation and is dependent on
   implementation and usage.  This calculation must be conservative.
   For example, a router sending packets into an IP tunnel must assume
   that the tunneled packets will not receive QoS control services
   unless it or the setup protocol can prove otherwise.

   Service-specific versions of the NON_IS_HOP flag indicate that one or
   more network elements along a path don't support the particular
   service. For example, the flag parameter identified by ID <2,2> being
   set indicates that some network element along the path does not
   support the Guaranteed service, though it might support another
   service such as Controlled-Load.

   If the global NON_IS_HOP flag <1,2> is set for a path, the receiver
   (network element or application) should consider the values of all
   other parameters defined in this specification, including service-
   specific NON_IS_HOP flags, as possibly inaccurate. If a service
   specific NON_IS_HOP flag is set for a path, the receiver should
   consider the values of all other parameters associated with that



Shenker/Wroclawski         Expires March, 1997                  [Page 7]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   service as possibly inaccurate.

   The NON_IS_HOP parameter may be represented in any form which can
   express boolean true and false. However, note that a network element
   must set this flag precisely when it does *not* fully understand the
   format or data representation of an arriving protocol message
   (because it does not support the specified service). Therefore, the
   data representation used for this parameter by setup and management
   protocols must allow the parameter value to be read and set even if
   the network element cannot otherwise parse the protocol message.

   An appropriate XDR description of this parameter is:

     bool NON_IS_HOP;

   However, the standard XDR data encoding for this description will not
   meet the requirement described above unless other restrictions are
   placed on message formats. An alternative data representation may be
   more appropriate.

     NOTE: The message format described for RSVP in [RFCRSVPUSE] carries
     this parameter as a single-bit flag, referred to as the "break
     bit".

 3.2 NUMBER_OF_IS_HOPS

   IS stands for "integrated services aware".  An integrated services
   aware network element is one that conforms to the various
   requirements described in this and other referenced documents.  The
   network element need not offer a specific service, but if it does it
   must support and characterize the service in conformance with the
   relevant specification, and if it does not it must correctly set the
   NON_IS_HOP flag parameter for the service. For completeness, the
   local parameter is assigned the parameter_number 3.

   The composition rule for this parameter is to increment the counter
   by one at each IS-aware hop.  This quantity, when composed end-to-
   end, informs the endpoint of the number of integrated-services aware
   network elements traversed along the path.  The parameter_number for
   this composed parameter is 4.

   Values of the composed parameter will range from 1 to 255, limited by
   the bound on IP hop count.

   The XDR representation of this parameter is:

     unsigned int NUMBER_OF_IS_HOPS;




Shenker/Wroclawski         Expires March, 1997                  [Page 8]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


 3.3. AVAILABLE_PATH_BANDWIDTH

   This parameter provides information about the bandwidth available
   along the path followed by a data flow.  The local parameter is an
   estimate of the bandwidth the network element has available for
   packets following the path.  Computation of the value of this
   parameter should take into account all information available to the
   network element about the path, taking into consideration
   administrative and policy controls on bandwidth, as well as physical
   resources.

      NOTE: This parameter should reflect, as closely as possible, the
      actual bandwidth available to packets following a path. However,
      the bandwidth available may depend on a number of factors not
      known to the network element until a specific QoS request is in
      place, such as the destination(s) of the packet flow, the service
      to be requested by the flow, or external policy information
      associated with a reservation request.  Because the parameter must
      in fact be provided before any specific QoS request is made, it is
      frequently difficult to provide the parameter accurately. In
      circumstances where the parameter cannot be provided accurately,
      the network element should make the best attempt possible, but is
      permitted to overestimate the available bandwidth by a significant
      amount.

   The parameter_number for AVAILABLE_PATH_BANDWIDTH is 5. The global
   parameter <1, 5> is an estimate of the bandwidth available to any
   packet following the path, without consideration of which (if any)
   QoS control service the packets may be subject to.

   In cases where a particular service is administratively or
   implementationally restricted to a limited portion of the overall
   available bandwidth, the service module may wish to export an
   override parameter which specifies this smaller bandwidth value.

   The composition rule for this parameter is the MIN function. The
   composed value is the minimum of the network element's value and the
   previously composed value. This quantity, when composed end-to-end,
   informs the endpoint of the minimal bandwidth link along the path
   from sender to receiver.  The parameter_number for the composed
   minimal bandwidth along the path is 6.

   Values of this parameter are measured in bytes per second.  The
   representation must be able to express values ranging from 1 byte per
   second to 40 terabytes per second, about what is believed to be the
   maximum theoretical bandwidth of a single strand of fiber.
   Particularly for large bandwidths, only the first few digits are
   significant, so the use of a floating point representation, accurate



Shenker/Wroclawski         Expires March, 1997                  [Page 9]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   to at least 0.1%, is encouraged.

   The XDR representation for this parameter is:

     float AVAILABLE_PATH_BANDWIDTH;

   For values of this parameter only valid non-negative floating point
   numbers are allowed. Negative numbers (including "negative zero"),
   infinities, and NAN's are not allowed.

      NOTE: An implementation which utilizes general-purpose hardware or
      software IEEE floating-point support may wish to verify that
      arriving parameter values meet these requirements before using the
      values in floating-point computations, in order to avoid
      unexpected exceptions or traps.

   If the network element cannot or wishes not to provide an estimate of
   path bandwidth, it may export a local value of zero for this
   parameter. A network element or application receiving a composed
   value of zero for this parameter should assume that the actual
   bandwidth available is unknown.

 3.4 MINIMUM_PATH_LATENCY

   The local parameter is the latency of the packet forwarding process
   associated with the network element, where the latency is defined to
   be the *minimal* possible packet delay added by the network element
   This delay may result from speed-of-light propagation delay, from
   packet processing limitations, or both. It does not include any
   variable queuing delay which may be present.

   The purpose of this parameter is to provide a baseline minimal path
   latency for use with services which provide estimates or bounds on
   additional path delay, such as Guaranteed [RFCG].  In conjunction
   with the queuing delay bound offered by Guaranteed and similar
   services, this parameter gives the application knowledge of both the
   minimum and maximum packet delivery delay.  Knowing both the minimum
   and maximum latency experienced by data packets allows the receiving
   application to accurately compute its de-jitter buffer requirements.

   Note that the quantity characterized by this parameter is the
   absolute lower bound on the packet processing and transmission
   latency of the network element. This value is the quantity required
   to provide jitter bounding. The parameter does *not* provide the type
   of upper-bound estimate on minimum latency which might be of interest
   for best-effort traffic and QoS control services which do not
   explicitly offer delay bounds. That is, the parameter will always
   underestimate, rather than overestimate, latency, particularly in



Shenker/Wroclawski         Expires March, 1997                 [Page 10]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   multicast and large cloud situations.

   When packets traversing a network element may experience different
   minimal latencies, this parameter should ideally report an accurate
   latency value for each path. For example, when an ATM point-
   multipoint virtual circuit is used to implement IP multicast, the
   mechanism used to implement this parameter would ideally compute a
   separate value for each destination. Doing so may require cooperation
   between the ingress and egress elements bounding the multi-path
   communication cloud.  The method by which this cooperation is
   achieved, and the choice of which IP-level network element actually
   provides and composes the value, is technology-dependent.

   An alternative choice is to provide the same value of this parameter
   for all paths through the cloud. The value reported must be the
   smallest latency for any possible path. Note that in this situation,
   QoS control services (e.g., Guaranteed) which provide an upper bound
   on latency cannot simply add their queuing delay to the value
   computed by this parameter; they must also compensate for path delays
   above the minimum. In this case the range between the minimum and
   maximum packet delays reported to the application will be larger,
   because the application will be told about the minimum delay along
   the shortest path and the maximum delay along the actual path.  This
   is acceptable in many situations.

   A third alternative is to report the "indeterminate" value, as
   specified below.  In this circumstance the client application may
   either deduce a minimum path latency through measurement, or assume a
   value of zero.

   The composition rule for this parameter is summation with a clamp of
   (2**32 - 1) on the maximum value. This quantity, when composed end-
   to-end, informs the endpoint of the minimal packet delay along the
   path from sender to receiver. The parameter_number for the latency of
   the network element's link is 7. The parameter_number for the
   cumulative latency along the path is 8.

   The latencies are measured in units of one microsecond. An individual
   element can advertise a latency value between 1 and 2**28 (somewhat
   over two minutes) and the total latency added across all elements can
   range as high as (2**32)-2. Should the sum of the different elements
   delay exceed (2**32)-2, the end-to-end advertised delay should be
   reported as indeterminate. This is described below.

   Note that while the granularity of measurement is microseconds, a
   conforming element is free to measure delays more loosely. The
   minimum requirement is that the element estimate its delay accurately
   to the nearest 100 microsecond granularity. Elements that can measure



Shenker/Wroclawski         Expires March, 1997                 [Page 11]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   more accurately are, of course, encouraged to do so.

      NOTE: Measuring in milliseconds is not acceptable, because if the
      minimum delay value is a millisecond, a path with several hops
      will lead to a composed delay of at least several milliseconds,
      which is likely to be misleading.

   The XDR description of this parameter is:

     unsigned int MINIMUM_PATH_LATENCY;

   The distinguished value (2**32)-1 is taken to mean "indeterminate
   latency". A network element which cannot accurately predict the
   latency of packets it is processing should set its local parameter to
   this value. Because the composition function limits the composed sum
   to this value, receipt of this value at a network element or
   application indicates that the true path latency is not known. This
   may happen because one or more network elements could not supply a
   value, or because the range of the composition calculation was
   exceeded.

 3.5. PATH_MTU

   This parameter computes the maximum transmission unit (MTU) for
   packets following a data path.  This value is required to invoke QoS
   control services which require that IP packet size be strictly
   limited to a specific MTU. Existing MTU discovery mechanisms cannot
   be used because they provide information only to the sender and they
   do not directly allow for QoS control services to specify MTU's
   smaller than the physical MTU.

   The local characterization parameter is the IP MTU, where the MTU of
   a network element is defined to be the maximum transmission unit the
   network element can accommodate without fragmentation, including IP
   and upper-layer protocol headers but not including link level
   headers.  The composition rule is to take the minimum of the network
   element's MTU and the previously composed value.  This quantity, when
   composed end-to-end, informs the endpoint of the maximum transmission
   unit that can traverse the path from sender to receiver without
   fragmentation.  The parameter_number for the MTU of the network
   element's link is 9.  The parameter_number for the composed MTU along
   the path is 10.

   A correct and valid value of this parameter must be provided by all
   IS-aware network elements.

   A specific service module may specify an MTU smaller than that of the
   overall network element by overriding this parameter with one giving



Shenker/Wroclawski         Expires March, 1997                 [Page 12]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   the service's MTU value. A service module may not specify an MTU
   value larger than that given by the global parameter.

   Values of this parameter are measured in bytes.  The representation
   must be able to express values ranging from 1 byte to 2**32-1 bytes.

   The XDR description of this parameter is:

     unsigned int PATH_MTU;

 3.6. TOKEN_BUCKET_TSPEC

   This parameter is used to describe data traffic parameters using a
   simple token bucket filter. This parameter is used by data senders to
   describe the traffic parameters of traffic it expects to generate,
   and by QoS control services to describe the parameters of traffic for
   which the reservation should apply. It is defined as a general rather
   than service-specific parameter because the same traffic description
   may be used by several QoS control services in some situations.

   The TOKEN_BUCKET_TSPEC parameter is assigned parameter_number 127.

   The TOKEN_BUCKET_TSPEC takes the form of a token bucket specification
   plus a peak rate p, minimum policed unit m, and a maximum packet size
   M.

   The token bucket specification includes an average or token rate r
   and a bucket depth b.  Both r and b must be positive.

   The token rate r is measured in bytes of IP datagrams per second.
   Values of this parameter may range from 1 byte per second to 40
   terabytes per second. In practice, only the first few digits of the r
   and p parameters are significant, so the use of floating point
   representations, accurate to at least 0.1% is encouraged.

   The bucket depth, b, is measured in bytes. Values of this parameter
   may range from 1 byte to 250 gigabytes. In practice, only the first
   few digits of the b parameter are significant, so the use of floating
   point representations, accurate to at least 0.1% is encouraged.

   The peak traffic rate p is measured in bytes of IP datagrams per
   second. Values of this parameter may range from 1 byte per second to
   40 terabytes per second. In practice, only the first few digits of
   the r and p parameters are significant, so the use of floating point
   representations, accurate to at least 0.1% is encouraged. The peak
   rate value may be set to positive infinity, indicating that it is
   unknown or unspecified.




Shenker/Wroclawski         Expires March, 1997                 [Page 13]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   The range of values allowed for these parameters is intentionally
   large to allow for future network technologies. Any particular
   network element is not expected to support the full range of values.

   The minimum policed unit, m, is an integer measured in bytes.  All IP
   datagrams less than size m will be counted against the token bucket
   as being of size m. The maximum packet size, M, is the biggest packet
   that will conform to the traffic specification; it is also measured
   in bytes.  Both m and M must be positive, and m must be less then or
   equal to M.

   The XDR description of this parameter is:

     struct {
       float r;
       float b;
       float p;
       unsigned m;
       unsigned M;
     } TOKEN_BUCKET_TSPEC;

   For the fields r and b only valid non-negative floating point numbers
   are allowed. Negative numbers (including "negative zero), infinities,
   and NAN's are not allowed.

   For the field p, only valid non-negative floating point numbers or
   positive infinity are allowed. Negative numbers (including "negative
   zero), negative infinities, and NAN's are not allowed.

      NOTE: An implementation which utilizes general-purpose hardware or
      software IEEE floating-point support may wish to verify that
      arriving parameter values meet these requirements before using the
      values in floating-point computations, in order to avoid
      unexpected exceptions or traps.

 4. Security Considerations

   Security considerations are not discussed in this memo.

5. References

   [RFCRSVP] B. Braden, et. al. "Resource Reservation Protocol (RSVP) -
   Version 1 Functional Specification", Internet Draft, July 1996,
   <draft-ietf-rsvp-spec-13.txt>

   [RFCRSVPUSE] J. Wroclawski. "The Use of RSVP with IETF Integrated
   Services", Internet Draft, October, 1996, <draft-ietf-intserv-rsvp-
   use-01.txt>



Shenker/Wroclawski         Expires March, 1997                 [Page 14]


INTERNET-DRAFT      draft-ietf-intserv-charac-02.txt       October, 1996


   [RFCTEMPLATE] S. Shenker and J. Wroclawski. "Network Element QoS
   Control Service Specification Template". Internet Draft, October
   1996, <draft-ietf-intserv-svc-template-03.txt>

   [RFCG] S. Shenker, C. Partridge, and R Guerin. "Specification of the
   Guaranteed Quality of Service", Internet Draft, July 1996, <draft-
   ietf-intserv-guaranteed-svc-06.txt>

   [RFCCL] J. Wroclawski. "Specification of the Controlled Load Quality
   of Service", Internet Draft, July 1996, <draft-ietf-intserv-ctrl-
   load-svc-03.txt>

   [RFCXDR] R. Srinivansan. "XDR: External Data Representation
   Standard", RFC 1832, August 1995.

Author's Address:

   Scott Shenker
   Xerox PARC
   3333 Coyote Hill Road
   Palo Alto, CA 94304-1314
   shenker@parc.xerox.com
   415-812-4840
   415-812-4471 (FAX)

   John Wroclawski
   MIT Laboratory for Computer Science
   545 Technology Sq.
   Cambridge, MA  02139
   jtw@lcs.mit.edu
   617-253-7885
   617-253-2673 (FAX)



















Shenker/Wroclawski         Expires March, 1997                 [Page 15]