| Delay-Tolerant Networking | E. Birrane |
| Internet-Draft | Johns Hopkins Applied Physics Laboratory |
| Intended status: Standards Track | March 11, 2019 |
| Expires: September 12, 2019 |
Asynchronous Management Protocol
draft-birrane-dtn-amp-06
This document describes a binary encoding of the Asynchronous Management Model (AMM) and a protocol for the exchange of these encoded items over a network. This Asynchronous Management Protocol (AMP) does not require transport-layer sessions, operates over unidirectional links, and seeks to reduce the energy and compute power necessary for performing network management on resource constrained devices.
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.
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."
This Internet-Draft will expire on September 12, 2019.
Copyright (c) 2019 IETF Trust and the persons identified as the document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
Network management in challenged and resource constrained networks must be accomplished differently than the network management methods in high-rate, high-availability networks. The Asynchronous Management Architecture (AMA) [I-D.birrane-dtn-ama] provides an overview and justification of an alternative to "synchronous" management services such as those provided by NETCONF. In particular, the AMA defines the need for a flexible, robust, and efficient autonomy engine to handle decisions when operators cannot be active in the network. The logical description of that autonomous model and its major components is given in the AMA Data Model (ADM) [I-D.birrane-dtn-adm].
The ADM presents an efficient and expressive autonomy model for the asynchronous management of a network node, but does not specify any particular encoding. This document, the Asynchronous Management Protocol (AMP), provides a binary encoding of AMM objects and specifies a protocol for the exchange of these encoded objects.
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 [RFC2119].
The AMP provides data monitoring, administration, and configuration for applications operating above the data link layer of the OSI networking model. While the AMP may be configured to support the management of network layer protocols, it also uses these protocol stacks to encapsulate and communicate its own messages.
It is assumed that the protocols used to carry AMP messages provide addressing, confidentiality, integrity, security, fragmentation/reassembly, and other network functions. Therefore, these items are outside of the scope of this document.
This document describes the format of messages used to exchange data models between managing and managed devices in a network. The rationale for this type of exchange is outside of the scope of this document and is covered in [I-D.birrane-dtn-ama]. The description and explanation of the data models exchanged is also outside of the scope of this document and is covered in [I-D.birrane-dtn-adm].
This document does not address specific configurations of AMP-enabled devices, nor does it discuss the interface between AMP and other management protocols.
Note: The terms "Actor", "Agent", "Application Data Model", "Externally Defined Data", "Variable", "Control", "Literal", "Macro", "Manager", "Report Template", "Report", "Table", "Constant", "Operator", "Time-Based Rule" and "State-Based Rule" are used without modification from the definitions provided in [I-D.birrane-dtn-ama].
The desirable properties of an asynchronous management protocol, as specified in the AMA, are summarized here to represent design constraints on the AMP specification.
+----------+----------+
| Field 1 | Field 2 |
| [TYPE 1] | [TYPE 2] |
| | (opt) |
+----------+----------+
Figure 1: Byte Field Formatting Example
The AMP specification provides an encoding of objects comprising the AMM. As such, AMP defines very few structures of its own. This section identifies those data structures that are unique to the AMP and required for it to perform appropriate and efficient encodings of AMM objects.
In the AMP, a "Nickname" (NN) is used to reduce the overall size of the encoding of ARIs that are defined in the context of an ADM. A NN is calculated as a function of an ADM Moderated Namespace and the type of object being identified.
As identifiers, ARIs are used heavily in AMM object definitions, particularly in those that define collections of objects. This makes encoding ARIs an important consideration when trying to optimize the size of AMP message.
Additionally, the majority of ARIs are defined in the context of an ADM. Certain AMM objects types (EDDs, OPs, CTRLs, TBLTs) can only be defined in the context of an ADM. Other object types (VARs, CONSTs, RPTTs) may have common, useful objects defined in an ADM as well. The structure of an ADM, to include its use of a Moderated Namespace and collections by object type, provide a regular structure that can be exploited for creating a compact representation.
In particular, as specified in [I-D.birrane-dtn-adm], ARIs can be grouped by (1) their namespace and (2) the type of AMM object being identified. For example, consider the following ARIs of type EDD defined in ADM1 with a Moderated Namespace of "/DTN/ADM1/".
ari:/DTN/ADM1/Edd.item_1 ari:/DTN/ADM1/Edd.item_2 ... ari:/DTN/ADM1/Edd.item_1974
In this case, the namespace (/DTN/ADM1/) and the object type (Edd) are good candidates for enumeration because their string encoding is very verbose and their information follows a regular structure shared across multiple ARIs. Separately, the string representation of object names (item_1, item_2, etc...) may be very verbose and they are also candidates for enumeration as they occupy a particular ADM object type in a particular order as published in the ADM.
Any ARI defined in an ADM exists in the context of a Moderated Namespace. These namespaces provide a unique string name for the ADM. However, ADMs can also be assigned a unique enumeration by the same moderating entities that ensure namespace uniqueness.
An ADM enumeration is an unsigned integer in the range of 0 to (2^64)/20. This range provides effective support for thousands of trillions of ADMs.
The formal set of ADMs, similar to SNMP MIBs and NETCONF YANG models, will be moderated and published. Additionally, a set of informal ADMs may be developed on a network-by-network or on an organization-by-organization bases.
Since informal ADMs exist within a predefined context (a network, an organization, or some other entity) they do not have individual ADM enumerations and are assigned the special enumeration "0". ARIs that are not defined in formal ADMs rely on other context information to help with their encoding (see Section 8.3).
An ADM Object Type Enumeration is an unsigned integer in the range of 0 - 19. This covers all of the standard areas for the ADM Template as defined in [I-D.birrane-dtn-adm]. Each of these types are enumerated in Table 1.
| AMM Object Type | Enumeration |
|---|---|
| CONST | 0 |
| CTRL | 1 |
| EDD | 2 |
| MAC | 3 |
| OPER | 4 |
| RPTT | 5 |
| SBR | 6 |
| TBLT | 7 |
| TBR | 8 |
| VAR | 9 |
| metadata | 10 |
| reserved | 11-19 |
As an enumeration, a Nickname is captured as a 64-bit unsigned integer (UVAST) calculated as a function of the ADM enumeration and the ADM type enumeration, as follows.
NN = ((ADM Enumeration) * 20) + (ADM Object Type Enumeration)
Considering the example set of ARIs from Section 7.1.1, assuming that ADM1 has ADM enumeration 9 and given that objects in the example were of type EDD, the NN for each of the 1974 items would be: (9 * 20) + 2 = 182. In this particular example, the ARI "/DTN/ADM1/Edd.item_1974" can be encoded in 5 bytes: two bytes to CBOR encode the nickname (182) and 3 bytes to CBOR encode the item's offset in the Edd collection (1974).
The assignment of formal ADM enumerations SHOULD take into consideration the nature of the applications and protocols to which the ADM applies. Those ADMs that are likely to be used in challenged networks SHOULD be allocated low enumeration numbers (e.g. those that will fit into 1-2 bytes) while ADMs that are likely to only be used in well resourced networks SHOULD be allocated higher enumeration numbers. It SHOULD NOT be the case that ADM enumerations are allocated on a first-come, first-served basis. It is recommended that ADM enumerations should be labeled based on the number of bytes of the Nickname as a function of the size of the ADM enumeration. These labels are shown in Table 2.
| ADM Enum | NN Size | Label | Comment |
|---|---|---|---|
| 0x1 - 0xCCC | 1-2 Bytes | Challenged Networks | Constraints imposed by physical layer and power. |
| 0xCCD - 0xCCCCCCC | 3-4 Bytes | Congested Networks | Constraints imposed by network traffic. |
| >=0xCCCCCCD | 5-8 Bytes | Resourced Networks | Generally unconstrained networks. |
This section describes the binary encoding of logical data constructs using the Concise Binary Object Representation (CBOR) defined in [RFC7049].
The following considerations act as guidance for CBOR encoders and decoders implementing the AMP.
The AMP encodes AMM primitive types as outlined in Table 3.
| AMM Type | CBOR Major Type | Comments |
|---|---|---|
| BYTE | unsigned int or byte string | BYTEs are individually encoded as unsigned integers unless the are defined as part of a byte string, in which case they are encoded as a single byte in the byte string. |
| INT | unsigned integer or negative integer | INTs are encoded as positive or negative integers from (u)int8_t up to (u)int32_t. |
| UINT | unsigned integer | UINTs are unsigned integers from uint8_t up to uint32_t. |
| VAST | unsigned integer or negative integer | VASTs are encoding as positive or negative integers up to (u)int64_t. |
| UVAST | unsigned integer | VASTs are unsigned integers up to uint64_t. |
| REAL32 | floating point | Up to an IEEE-754 Single Precision Float. |
| REAL64 | floating point | Up to an IEEE-754 Double Precision Float. |
| STRING | text string | Uses CBOR encoding unmodified. |
| BOOL | Simple Value | 0 is considered FALSE. 1 is considered TRUE. |
This section provides the CBOR encodings for AMM derived types.
The AMM derived type Byte String (BYTESTR) is encoded as a CBOR byte string.
An TV is encoded as a UVAST. Similarly, a TS is also encoded as a UVAST since a TS is simply an absolute TV.
Rather than define two separate encodings for TVs (one for absolute TVs and one for relative TVs) a single, unambiguous encoding can be generated by defining a Relative Time Epoch (RTE) and interpreting the type of TV in relation to that epoch. Time values less than the RTE MUST be interpreted as relative times. Time values greater than or equal to the RTE MUST be interpreted as absolute time values.
A relative TV is encoded as the number of seconds after an initiating event. An absolute TV (and TS) is encoded as the number of seconds that have elapsed since 1 Jan 2000 00:00:00 (Unix Time 946684800).
The RTE is defined as the timestamp value for September 9th, 2017 (Unix time 1504915200). Since TS values are the number of seconds since 1 Jan 2000 00:00:00, the RTE as a TS value is 1504915200 - 946684800 = 558230400.
Potential Time values
________________________/\________________________
/ \
Relative Times Absolute Times
<------------------------><------------------------>
0 - 558,230,400 558,230,401 - 2^64
|------------------------|-------------------------|
| |
00:00:00 1 Jan 2000 00:00:00 9 Sep 2017
Unix Time 946684800 Unix Time 1504915200
The potential values of TV, and how they should be interpreted as relative or absolute is illustrated below.
For example, a time value of "10" is a relative time representing 10 seconds after an initiating event. A time value of "600,000,000" refers to Saturday, 5 Jan, 2019 10:40:00.
IF (time_value <= 558230400) THEN
absolute_time = (event_time - 946684800) + time_value
ELSE
absolute_time = time_value
NOTE: Absolute and relative times are interchangeable. An absolute time can be converted into a relative time by subtracting the current time from the absolute time. A relative time can be converted into an absolute time by adding to the relative time the timestamp of its relative event. A pseudo-code example of converting a relative time to an absolute time is as follows, assuming that current-time is expressed in Unix Epoch time.
+---------+
| TNV |
| [ARRAY] |
+----++---+
||
||
_______________/ \________________
/ \
+------------+-----------+----------+
| Flags/Type | Name | Value |
| [BYTE] | [TXT STR] | [Varies] |
| | (opt) | (opt) |
+------------+-----------+----------+
Figure 2: E(TNV) Format
TNV values are encoded as a CBOR array that comprises four distinct pieces of information: a set of flags, a type, an optional name, and an optional value. In the E(TNV) the flag and type information are compressed into a single value. The CBOR array MUST have length 1, 2, or 3 depending on the number of optional fields appearing in the encoding. The E(TNV) format is illustrated in Figure 2.
The E(TNV) fields are defined as follows.
E(TNV) Flag/Type Byte Format
+------+---------------+
| Name | Struct |
| Flag | Type |
+------+---------------+
| 7 | 6 5 4 3 2 1 0 |
+------+---------------+
MSB LSB
Figure 3
A TNV Collection (TNVC) is an ordered set of TNVs with special semantics for more efficiently encoding sets of TNVs with identical attributes.
A TNV, defined in Section 8.2.2.3, consists of three distinct components: a type, a name, and a value. When all of the TNVs in the TNVC have the same format (such as they all include type information) then the encoding of the TNVC can use this information to save encoding space and make processing more efficient. In cases when all TNVs have the same format, the types (if present), names (if present), and values (if present) are separated into their own arrays for individual processing with type information (if present) always appearing first.
Extracting type information to the "front" of the collection optimizes the performance of type validators. A validator can inspect the first array to ensure that element values match type expectations. If type information were distributed throughout the collection, as in the case with the TNVC, a type validator would need to scan through the entire set of data to validate each type in the collection.
+----------+
| TNVC |
| [ARRAY] |
+----++----+
||
||
______________________/ \_______________________
/ \
+--------+-----------+---------+---------+---------+
| Flags | Types | Names | Values | Mixed |
| [BYTE] | [BYTESTR] | [ARRAY] | [ARRAY] | [ARRAY] |
| | (Opt) | (Opt) | (Opt) | (Opt) |
+--------+-----------+---------+---------+---------+
Figure 4: E(TNVC) Format
A TNVC is encoded as a CBOR array with at least one element, the flags which represent the optional portions of the collection that are present. If the TNVC represents an empty set, there will be no additional entries. The format of a TNVC is illustrated in Figure 4.
The E(TNVC) fields are defined as follows.
E(TNV) Flag Byte Format
+----------+------+------+------+------+
| Reserved | Mix | Type | Name | Val |
| Flags | Flag | Flag | Flag | Flag |
+----------+------+------+------+------+
| 7-4 | 3 | 2 | 1 | 0 |
+----------+------+------+------+------+
MSB LSB
Figure 5
An ARI collection is an ordered collection of ARI values. It is encoded as a CBOR array with each element being an encoded ARI, as illustrated in Figure 6.
E(AC) Format
+---------+
| AC |
| [ARRAY] |
+----++---+
||
||
________/ \_________
/ \
+-------+ +-------+
| ARI 1 | ... | ARI N |
| [ARI] | | [ARI] |
+-------+ +-------+
Figure 6
The Expression object encapsulates a typed postfix expression in which each operator MUST be of type OPER and each operand MUST be the typed result of an operator or one of EDD, VAR, LIT, or CONST.
The Expression object is encoded as a CBOR byte string whose format is illustrated in Figure 7.
E(EXPR) Format
+--------+------------+
| Type | Expression |
| [BYTE] | [AC] |
+--------+------------+
Figure 7
The ARI, as defined in [I-D.birrane-dtn-adm], identifies an AMM object. There are two kinds of objects that can be identified in this scheme: literal objects (of type LIT) and all other objects.
A literal identifier is one that is literally defined by its value, such as numbers (0, 3.14) and strings ("example"). ARIs of type LITERAL do not have issuers or nicknames or parameters. They are simply typed basic values.
The E(ARI) of a LIT object is encoded as a CBOR byte string and consists of a mandatory flag BYTE and the value of the LIT.
The E(ARI) structure for LIT types is illustrated in Figure 8.
E(ARI) Literal Format
+--------+----------+
| Flags | Value |
| [BYTE] | [VARIES] |
+--------+----------+
Figure 8
These fields are defined as follows.
E(ARI) Literal Flag Byte Format
+------------+-------------+
| VALUE TYPE | STRUCT TYPE |
+------------+-------------|
| 7 6 5 4 | 3 2 1 0 |
+------------+-------------+
MSB LSB
Figure 9
All other ARIs are defined in the context of AMM objects and may contain parameters and other meta-data. The AMP, as a machine-to-machine binary encoding of this information removes human-readable information such as Name and Description from the E(ARI). Additionally, this encoding adds other information to improve the efficiency of the encoding, such as the concept of Nicknames, defined in Section 7.1.
The E(ARI) is encoded as a CBOR byte string and consists of a mandatory flag byte, an encoded object name, and optional annotations to assist with filtering, access control, and parameterization. The E(ARI) structure is illustrated in Figure 10.
E(ARI) General Format
+--------+---------+-----------+---------+---------+-----------+
| Flags | NN | Name | Parms | Issuer | Tag |
| [BYTE] | [UVAST] | [BYTESTR] | [TNVC] | [UVAST] | [BYTESTR] |
| | (opt) | | (opt) | (opt) | opt) |
+--------+---------+-----------+---------+---------+-----------+
Figure 10
These fields are defined as follows.
E(ARI) General Flag Byte Format
+----+------+-----+-----+-------------+
| NN | PARM | ISS | TAG | STRUCT TYPE |
+----+------+-----+-----+-------------+
| 7 | 6 | 5 | 4 | 3 2 1 0 |
+----+------+-----+-----+-------------+
MSB LSB
Figure 11
The autonomy model codified in [I-D.birrane-dtn-adm] comprises multiple individual objects. This section describes the CBOR encoding of these objects.
Note: The encoding of an object refers to the way in which the complete object can be encoded such that the object as it exists on a Manager may be re-created on an Agent, and vice-versa. In cases where both a Manager and an Agent already have the definition of an object, then only the encoded ARI of the object needs to be communicated. This is the case for all objects defined in a published ADM and any user-defined object that has been synchronized between an Agent and Manager.
Externally defined data (EDD) are solely defined in the ADMs for various applications and protocols. EDDs represent values that are calculated external to an AMA Agent, such as values measured by firmware.
The representation of these data is simply their identifying ARIs. The representation of an EDD is illustrated in Figure 12.
E(EDD) Format
+-------+
| ID |
| [ARI] |
+-------+
Figure 12
Unlike Literals, a Constant is an immutable, typed, named value. Examples of constants include PI to some number of digits or the UNIX Epoch.
Since ADM definitions are preconfigured on Agents and Managers in an AMA, the type information for a given Constant is known by all actors in the system and the encoding of the Constant needs to only be the name of the constant as the Manager and Agent can derive the type and value from the unique Constant name.
The format of a Constant is illustrated in Figure 13.
E(CONST) Format
+-------+
| ID |
| [ARI] |
+-------+
Figure 13
A Control represents a pre-defined and optionally parameterized opcode that can be run on an Agent. Controls in the AMP are always defined in the context of an AMA; there is no concept of an operator-defined Control. Since Controls are pre-configured in Agents and Managers as part of ADM support, their representation is the ARI that identifies them, similar to EDDs.
The format of a Control is illustrated in Figure 14.
E(CTRL) Format
+-------+
| ID |
| [ARI] |
+-------+
Figure 14
Macros in the AMP are ordered collections of ARIs (an AC) that contain Controls or other Macros. When run by an Agent, each ARI in the AC MUST be run in order.
Any AMP implementation MUST allow at least 4 levels of Macro nesting. Implementations MUST prevent recursive nesting of Macros.
The ARI associated with a Macro MAY contain parameters. Each parameter present in a Macro ARI MUST contain type, name, and value information. Any Control or Macro encapsulated within a parameterized Macro MAY also contain parameters. If an encapsulated object parameter contains only name information, then the parameter value MUST be taken from the named parameter provided by the encapsulating Macro. Otherwise, the value provided to the object MUST be used instead.
The format of a Macro is illustrated in Figure 15.
E(MAC) Format
+-------+------------+
| ID | Definition |
| [ARI] | [AC] |
+-------+------------+
Figure 15
Operators are always defined in the context of an ADM. There is no concept of a user-defined operator, as operators represent mathematical functions implemented by the firmware on an Agent. Since Operators are preconfigured in Agents and Managers as part of ADM support, their representation is simply the ARI that identifies them.
The ADM definition of an Operator MUST specify how many parameters are expected and the expected type of each parameter. For example, the unary NOT Operator ("!") would accept one parameter. The binary PLUS Operator ("+") would accept two parameters. A custom function to calculate the average of the last 10 samples of a data item should accept 10 parameters.
Operators are always evaluated in the context of an Expression. The encoding of an Operator is illustrated in Figure 16.
E(OP) Format
+-------+
| ID |
| [ARI] |
+-------+
Figure 16
A Report Template is an ordered collection of identifiers that describe the order and format of data in any Report built in compliance with the template. A template is a schema for a class of reports. It contains no actual values and may be defined in a formal ADM or configured by users in the context of a network deployment.
The encoding of a RPTT is illustrated in Figure 17.
E(RPTT) Format
+-------+----------+
| ID | Contents |
| [ARI] | [AC] |
+-------+----------+
Figure 17
A Report is a set of data values populated using a given Report Template. While Reports do not contain name information, they MAY contain type information in cases where recipients wish to perform type validation prior to interpreting the Report contents in the context of a Report Template. Reports are "anonymous" in the sense that any individual Report does not contain a unique identifier. Reports can be differentiated by examining the combination of (1) the Report Template being populated, (2) the time at which the Report was populated, and (3) the Agent producing the report.
A Report object is comprised of the identifier of the template used to populate the report, an optional timestamp, and the contents of the report. A Report is encoded as a CBOR array with either 2 or 3 elements. If the array has 2 elements then the optional Timestamp MUST NOT be in the Report encoding. If the array has 3 elements then the optional timestamp MUST be included in the Report encoding. The Report encoding is illustrated in Figure 18.
E(RPT) Format
+---------+
| RPT |
| [ARRAY] |
+---++----+
||
||
_____________/ \_____________
/ \
+----------+-----------+---------+
| Template | Timestamp | Entries |
| [ARI] | [TS] | [TNVC] |
| | (opt) | |
+----------+-----------+---------+
Figure 18
A State-Based Rule (SBR) specifies that a particular action should be taken by an Agent based on some evaluation of the internal state of the Agent. A SBR specifies that starting at a particular START time an ACTION should be run by the Agent if some CONDITION evaluates to true, until the ACTION has been run COUNT times. When the SBR is no longer valid it may be discarded by the agent.
Examples of SBRs include:
An SBR object is encoded as a CBOR array with 5 elements as illustrated in Figure 19.
E(SBR) Format
+---------+
| SBR |
| [ARRAY] |
+---++----+
||
||
_______________________/ \_______________________
/ \
+-------+-------+--------+--------+--------+--------+
| ID | START | COND | EVALS | FIRES | ACTION |
| [ARI] | [TV] | [EXPR] | [UINT] | [UINT] | [AC] |
+-------+-------+--------+--------+--------+--------+
Figure 19
A Table Template (TBLT) describes the types, and optionally names, of the columns that define a Table.
Because TBLTs are only defined in the context of an ADM, their definition cannot change operationally. Therefore, a TBLT is encoded simply as the ARI for the template. The format of the TBLT Object Array is illustrated in Figure 20.
E(TBLT) Format
+-------+
| ID |
| [ARI] |
+-------+
Figure 20
The elements of the TBLT array are defined as follows.
A Table object describes the series of values associated with a Table Template.
A Table object is encoded as a CBOR array, with the first element of the array identifying the Table Template and each subsequent element identifying a row in the table. The format of the TBL Object Array is illustrated in Figure 21.
E(TBL) Format
+---------+
| TBL |
| [ARRAY] |
+---++----+
||
||
______________/ \_______________
/ \
+---------+--------+ +--------+
| TBLT ID | Row 1 | | Row N |
| [ARI] | [TNVC] | ... | [TNVC] |
+---------+--------+ +--------+
Figure 21
The TBL fields are defined as follows.
A Time-Based Rule (TBR) specifies that a particular action should be taken by an Agent based on some time interval. A TBR specifies that starting at a particular START time, and for every PERIOD seconds thereafter, an ACTION should be run by the Agent until the ACTION has been run for COUNT times. When the TBR is no longer valid it MAY BE discarded by the Agent.
Examples of TBRs include:
The TBR object is encoded as a CBOR array with 5 elements as illustrated in Figure 22.
E(TBR) Format
+---------+
| TBR |
| [ARRAY] |
+---++----+
||
||
___________________/ \___________________
/ \
+-------+-------+--------+--------+--------+
| ID | START | PERIOD | COUNT | ACTION |
| [ARI] | [TV] | [UINT] | [UINT] | [AC] |
+-------+-------+--------+--------+--------+
Figure 22
Variable objects are transmitted in the AMP without the human-readable description.
Variable objects are encoded as a CBOR byte string whose format is illustrated in Figure 23.
E(VAR) Format
+------------+
| Variable |
| [BYTESTR] |
+-----++-----+
||
||
______/ \_____
/ \
+-------+-------+
| ID | Value |
| [ARI] | [TNV] |
+-------+-------+
Figure 23
This section describes the format of the messages that comprise the AMP protocol.
The AMP message specification is limited to three basic communications:
| Message | Enumeration | Description |
|---|---|---|
| Register Agent | 0 | Add Agents to the list of managed devices known to a Manager. |
| Report Set | 1 | Receiving a Report of one or more Report Entries from an Agent. |
| Perform Control | 2 | Sending a Macro of one or more Controls to an Agent. |
The entire management of a network can be performed using these three messages and the configurations from associated ADMs.
Individual messages within the AMP are combined into a single group for communication with another AMP Actor. Messages within a group MUST be received and applied as an atomic unit. The format of a message group is illustrated in Figure 24. These message groups are assumed communicated amongst Agents and Managers as the payloads of encapsulating protocols which should provide additional security and data integrity features as needed.
A message group is encoded as a CBOR array with at least 2 elements, the first being the time the group was created followed by 1 or more messages that comprise the group. The format of the message group is illustrated in Figure 24.
AMP Message Group Format
+---------------+
| Message Group |
| [ARRAY] |
+------++-------+
||
____________________||___________________
/ \
+-----------+-----------+ +-----------+
| Timestamp | Message 1 | ... | Message N |
| [TS] | [BYTESTR] | | [BYTESTR] |
+-----------+-----------+ +-----------+
Figure 24
Each message identified in the AMP specification adheres to a common message format, illustrated in Figure 25, consisting of a message header, a message body, and an optional trailer.
Each message in the AMP is encode as a CBOR byte string formatted in accordance with Figure 25.
AMP Message Format
+--------+----------+----------+
| Header | Body | Trailer |
| [BYTE] | [VARIES] | [VARIES] |
| | | (opt.) |
+--------+----------+----------+
Figure 25
AMP Common Message Header
+----------+-----+------+-----+----------+
| Reserved | ACL | Nack | Ack | Opcode |
+----------+-----+------+-----+----------+
| 7 6 | 5 | 4 | 3 | 2 1 0 |
+----------+-----+------+-----+----------+
MSB LSB
Figure 26
The Register Agent message is used to inform an AMP Manager of the presence of another Agent in the network.
The body of this message is the name of the new agent, encoded as illustrated in Figure 27.
Register Agent Message Body
+-----------+
| Agent ID |
| [BYTESTR] |
+-----------+
Figure 27
The Report Set message contains a set of 1 or more Reports produced by an AMP Agent and sent to an AMP Manager.
The body of this message contains information on the recipient of the reports followed by one or more Reports. The body is encoded as illustrated in Figure 28.
Report Set Message Body
+----------+--------+ +--------+
| RX Names | RPT 1 | | RPT N |
| [ARRAY] | [RPT] | ... | [RPT] |
+----------+--------+ +--------+
Figure 28
The perform control message causes the receiving AMP Actor to run one or more preconfigured Controls provided in the message.
The body of this message is the start time for the controls followed by the controls themselves, as illustrated in Figure 29.
Perform Control Message Body
+-------+-----------+
| Start | Controls |
| [TV] | [AC] |
+-------+-----------+
Figure 29
A Nickname registry needs to be established.
Security within the AMP exists in two layers: transport layer security and access control.
Transport-layer security addresses the questions of authentication, integrity, and confidentiality associated with the transport of messages between and amongst Managers and Agents. This security is applied before any particular Actor in the system receives data and, therefore, is outside of the scope of this document.
Finer grain application security is done via ACLs provided in the AMP message headers.
A reference implementation of this version of the AMP specification is available in the 3.6.2 release of the ION open source code base available from sourceforge at https://sourceforge.net/projects/ion-dtn/.
| [I-D.birrane-dtn-ama] | Birrane, E., "Asynchronous Management Architecture", Internet-Draft draft-birrane-dtn-ama-07, June 2018. |
| [I-D.birrane-dtn-adm] | Birrane, E., DiPietro, E. and D. Linko, "AMA Application Data Model", Internet-Draft draft-birrane-dtn-adm-02, June 2018. |
| [RFC2119] | Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997. |
| [RFC7049] | Bormann, C. and P. Hoffman, "Concise Binary Object Representation (CBOR)", RFC 7049, DOI 10.17487/RFC7049, October 2013. |
The following participants contributed technical material, use cases, and useful thoughts on the overall approach to this protocol specification: Jeremy Pierce-Mayer of INSYEN AG contributed the concept of the typed data collection and early type checking in the protocol. David Linko and Evana DiPietro of the Johns Hopkins University Applied Physics Laboratory contributed appreciated review and type checking of various elements of this specification.