Network Working Group S. Waldbusser Request for Comments: 2021 INS Category: Standards Track January 1997 Remote Network Monitoring Management Information Base Version 2 using SMIv2 Status of this Memo This document specifies an Internet standards track protocol for the Internet community, and requests discussion and suggestions for improvements. Please refer to the current edition of the "Internet Official Protocol Standards" (STD 1) for the standardization state and status of this protocol. Distribution of this memo is unlimited. Abstract This memo defines a portion of the Management Information Base (MIB) for use with network management protocols in TCP/IP-based internets. In particular, it defines objects for managing remote network monitoring devices. Table of Contents 1 The Network Management Framework ..................... 2 2 Overview ............................................. 2 2.1 Remote Network Management Goals .................... 3 2.2 Structure of MIB ................................... 5 3 Control of Remote Network Monitoring Devices ......... 6 3.1 Resource Sharing Among Multiple Management Sta- tions ............................................. 7 3.2 Row Addition Among Multiple Management Stations .... 9 4 Conventions .......................................... 10 5 RMON 2 Conventions ................................... 10 5.1 Usage of the term Application Level ................ 10 5.2 Protocol Directory and Limited Extensibility ....... 11 5.3 Errors in packets .................................. 11 6 Definitions .......................................... 12 7 Security Considerations .............................. 122 8 Appendix - TimeFilter Implementation Notes .......... 123 9 Acknowledgments ..................................... 129 10 References .......................................... 129 11 Author's Address..................................... 130
1. The Network Management Framework The Internet-standard Network Management Framework consists of three components. They are: RFC 1902  which defines the SMI, the mechanisms used for describing and naming objects for the purpose of management. RFC 1213, STD 17,  which defines MIB-II, the core set of managed objects for the Internet suite of protocols. RFC 1905  which defines the SNMP, the protocol used for network access to managed objects. The Framework permits new objects to be defined for the purpose of experimentation and evaluation. Managed objects are accessed via a virtual information store, termed the Management Information Base or MIB. Within a given MIB module, objects are defined using the SMI's OBJECT-TYPE macro. At a minimum, each object has a name, a syntax, an access-level, and an implementation-status. The name is an object identifier, an administratively assigned name, which specifies an object type. The object type together with an object instance serves to uniquely identify a specific instantiation of the object. For human convenience, we often use a textual string, termed the object descriptor, to also refer to the object type. The syntax of an object type defines the abstract data structure corresponding to that object type. The ASN.1  language is used for this purpose. However, RFC 1902 purposely restricts the ASN.1 constructs which may be used. These restrictions are explicitly made for simplicity. The access-level of an object type defines whether it makes "protocol sense" to read and/or write the value of an instance of the object type. (This access-level is independent of any administrative authorization policy.) The implementation-status of an object type indicates whether the object is mandatory, optional, obsolete, or deprecated. 2. Overview This document continues the architecture created in the RMON MIB [RFC 1757] by providing a major feature upgrade, primarily by providing RMON analysis up to the application layer.
Remote network monitoring devices, often called monitors or probes, are instruments that exist for the purpose of managing a network. Often these remote probes are stand-alone devices and devote significant internal resources for the sole purpose of managing a network. An organization may employ many of these devices, one per network segment, to manage its internet. In addition, these devices may be used for a network management service provider to access a client network, often geographically remote. The objects defined in this document are intended as an interface between an RMON agent and an RMON management application and are not intended for direct manipulation by humans. While some users may tolerate the direct display of some of these objects, few will tolerate the complexity of manually manipulating objects to accomplish row creation. These functions should be handled by the management application. 2.1. Remote Network Management Goals o Offline Operation There are sometimes conditions when a management station will not be in constant contact with its remote monitoring devices. This is sometimes by design in an attempt to lower communications costs (especially when communicating over a WAN or dialup link), or by accident as network failures affect the communications between the management station and the probe. For this reason, this MIB allows a probe to be configured to perform diagnostics and to collect statistics continuously, even when communication with the management station may not be possible or efficient. The probe may then attempt to notify the management station when an exceptional condition occurs. Thus, even in circumstances where communication between management station and probe is not continuous, fault, performance, and configuration information may be continuously accumulated and communicated to the management station conveniently and efficiently.
o Proactive Monitoring Given the resources available on the monitor, it is potentially helpful for it continuously to run diagnostics and to log network performance. The monitor is always available at the onset of any failure. It can notify the management station of the failure and can store historical statistical information about the failure. This historical information can be played back by the management station in an attempt to perform further diagnosis into the cause of the problem. o Problem Detection and Reporting The monitor can be configured to recognize conditions, most notably error conditions, and continuously to check for them. When one of these conditions occurs, the event may be logged, and management stations may be notified in a number of ways. o Value Added Data Because a remote monitoring device represents a network resource dedicated exclusively to network management functions, and because it is located directly on the monitored portion of the network, the remote network monitoring device has the opportunity to add significant value to the data it collects. For instance, by highlighting those hosts on the network that generate the most traffic or errors, the probe can give the management station precisely the information it needs to solve a class of problems. o Multiple Managers An organization may have multiple management stations for different units of the organization, for different functions (e.g. engineering and operations), and in an attempt to provide disaster recovery. Because environments with multiple management stations are common, the remote network monitoring device has to deal with more than own management station, potentially using its resources concurrently.
2.2. Structure of MIB The objects are arranged into the following groups: - protocol directory - protocol distribution - address mapping - network layer host - network layer matrix - application layer host - application layer matrix - user history - probe configuration These groups are the basic units of conformance. If a remote monitoring device implements a group, then it must implement all objects in that group. For example, a managed agent that implements the network layer matrix group must implement the nlMatrixSDTable and the nlMatrixDSTable. Implementations of this MIB must also implement the system and interfaces group of MIB-II . MIB-II may also mandate the implementation of additional groups. These groups are defined to provide a means of assigning object identifiers, and to provide a method for managed agents to know which objects they must implement. This document also contains enhancements to tables defined in the RMON MIB [RFC 1757]. These enhancements include: 1) Adding the DroppedFrames and LastCreateTime conventions to each table defined in the RMON MIB. 2) Augmenting the RMON filter table with a mechanism that allows filtering based on an offset from the beginning of a particular protocol, even if the protocol headers are variable length.
3) Augmenting the RMON filter and capture status bits with additional bits for WAN media and generic media. These bits are defined here as: Bit Definition 6 For WAN media, this bit is set for packets coming from one direction and cleared for packets coming from the other direction. It is an implementation specific matter as to which bit is assigned to which direction, but it must be consistent for all packets received by the agent, and if the agent knows which end of the link is "local" and which end is "network", the bit should be set for packets from the "local" side and should be cleared for packets from the "network" side. 7 For any media, this bit is set for any packet with a physical layer error. This bit may be set in addition to other media-specific bits that denote the same condition. 8 For any media, this bit is set for any packet that is too short for the media. This bit may be set in addition to other media-specific bits that denote the same condition. 9 For any media, this bit is set for any packet that is too long for the media. This bit may be set in addition to other media-specific bits that denote the same condition. These enhancements are implemented by RMON-2 probes that also implement RMON and do not add any requirements to probes that are compliant to just RMON. 3. Control of Remote Network Monitoring Devices Due to the complex nature of the available functions in these devices, the functions often need user configuration. In many cases, the function requires parameters to be set up for a data collection operation. The operation can proceed only after these parameters are fully set up. Many functional groups in this MIB have one or more tables in which to set up control parameters, and one or more data tables in which to place the results of the operation. The control tables are typically read/write in nature, while the data tables are typically read/only.
Because the parameters in the control table often describe resulting data in the data table, many of the parameters can be modified only when the control entry is not active. Thus, the method for modifying these parameters is to de-activate the entry, perform the SNMP Set operations to modify the entry, and then re-activate the entry. Deleting the control entry causes the deletion of any associated data entries, which also gives a convenient method for reclaiming the resources used by the associated data. Some objects in this MIB provide a mechanism to execute an action on the remote monitoring device. These objects may execute an action as a result of a change in the state of the object. For those objects in this MIB, a request to set an object to the same value as it currently holds would thus cause no action to occur. To facilitate control by multiple managers, resources have to be shared among the managers. These resources are typically the memory and computation resources that a function requires. 3.1. Resource Sharing Among Multiple Management Stations When multiple management stations wish to use functions that compete for a finite amount of resources on a device, a method to facilitate this sharing of resources is required. Potential conflicts include: o Two management stations wish to simultaneously use resources that together would exceed the capability of the device. o A management station uses a significant amount of resources for a long period of time. o A management station uses resources and then crashes, forgetting to free the resources so others may use them. The OwnerString mechanism is provided for each management station initiated function in this MIB to avoid these conflicts and to help resolve them when they occur. Each function has a label identifying the initiator (owner) of the function. This label is set by the initiator to provide for the following possibilities: o A management station may recognize resources it owns and no longer needs. o A network operator can find the management station that owns the resource and negotiate for it to be freed. o A network operator may decide to unilaterally free resources another network operator has reserved.
o Upon initialization, a management station may recognize resources it had reserved in the past. With this information it may free the resources if it no longer needs them. Management stations and probes should support any format of the owner string dictated by the local policy of the organization. It is suggested that this name contain one or more of the following: IP address, management station name, network manager's name, location, or phone number. This information will help users to share the resources more effectively. There is often default functionality that the device or the administrator of the probe (often the network administrator) wishes to set up. The resources associated with this functionality are then owned by the device itself or by the network administrator, and are intended to be long-lived. In this case, the device or the administrator will set the relevant owner object to a string starting with 'monitor'. Indiscriminate modification of the monitor-owned configuration by network management stations is discouraged. In fact, a network management station should only modify these objects under the direction of the administrator of the probe. Resources on a probe are scarce and are typically allocated when control rows are created by an application. Since many applications may be using a probe simultaneously, indiscriminate allocation of resources to particular applications is very likely to cause resource shortages in the probe. When a network management station wishes to utilize a function in a monitor, it is encouraged to first scan the control table of that function to find an instance with similar parameters to share. This is especially true for those instances owned by the monitor, which can be assumed to change infrequently. If a management station decides to share an instance owned by another management station, it should understand that the management station that owns the instance may indiscriminately modify or delete it. It should be noted that a management application should have the most trust in a monitor-owned row because it should be changed very infrequently. A row owned by the management application is less long-lived because a network administrator is more likely to re- assign resources from a row that is in use by one user than from a monitor-owned row that is potentially in use by many users. A row owned by another application would be even less long-lived because the other application may delete or modify that row completely at its discretion.
3.2. Row Addition Among Multiple Management Stations The addition of new rows is achieved using the RowStatus method described in RFC 1903 . In this MIB, rows are often added to a table in order to configure a function. This configuration usually involves parameters that control the operation of the function. The agent must check these parameters to make sure they are appropriate given restrictions defined in this MIB as well as any implementation specific restrictions such as lack of resources. The agent implementor may be confused as to when to check these parameters and when to signal to the management station that the parameters are invalid. There are two opportunities: o When the management station sets each parameter object. o When the management station sets the row status object to active. If the latter is chosen, it would be unclear to the management station which of the several parameters was invalid and caused the badValue error to be emitted. Thus, wherever possible, the implementor should choose the former as it will provide more information to the management station. A problem can arise when multiple management stations attempt to set configuration information simultaneously using SNMP. When this involves the addition of a new conceptual row in the same control table, the managers may collide, attempting to create the same entry. To guard against these collisions, each such control entry contains a status object with special semantics that help to arbitrate among the managers. If an attempt is made with the row addition mechanism to create such a status object and that object already exists, an error is returned. When more than one manager simultaneously attempts to create the same conceptual row, only the first will succeed. The others will receive an error. In the RMON MIB [RFC 1757], the EntryStatus textual convention was introduced to provide this mutual exclusion function. Since then, this function was added to the SNMP framework as the RowStatus textual convention. The RowStatus textual convention is used for the definition of all new tables. When a manager wishes to create a new control entry, it needs to choose an index for that row. It may choose this index in a variety of ways, hopefully minimizing the chances that the index is in use by another manager. If the index is in use, the mechanism mentioned previously will guard against collisions. Examples of schemes to choose index values include random selection or scanning the control
table looking for the first unused index. Because index values may be any valid value in the range and they are chosen by the manager, the agent must allow a row to be created with any unused index value if it has the resources to create a new row. Some tables in this MIB reference other tables within this MIB. When creating or deleting entries in these tables, it is generally allowable for dangling references to exist. There is no defined order for creating or deleting entries in these tables. 4. Conventions The following conventions are used throughout the RMON MIB and its companion documents. Good Packets Good packets are error-free packets that have a valid frame length. For example, on Ethernet, good packets are error-free packets that are between 64 octets long and 1518 octets long. They follow the form defined in IEEE 802.3 section 3.2.all. Bad Packets Bad packets are packets that have proper framing and are therefore recognized as packets, but contain errors within the packet or have an invalid length. For example, on Ethernet, bad packets have a valid preamble and SFD, but have a bad CRC, or are either shorter than 64 octets or longer than 1518 octets. 5. RMON 2 Conventions The following practices and conventions are introduced in the RMON 2 MIB. 5.1. Usage of the term Application Level There are many cases in this MIB where the term Application Level is used to describe a class of protocols or a capability. This does not typically mean a protocol that is an OSI Layer 7 protocol. Rather, it is used to identify a class of protocols that is not limited to MAC-layer and network-layer protocols, but can also include transport, session, presentation, and application-layer protocols.
5.2. Protocol Directory and Limited Extensibility Every RMON 2 implementation will have the capability to parse certain types of packets and identify their protocol type at multiple levels, The protocol directory presents an inventory of those protocol types the probe is capable of monitoring, and allows the addition, deletion, and configuration of protocol types in this list. One concept deserves special attention: the "limited extensibility" of the protocol directory table. The RMON 2 model is that protocols are detected by static software that has been written at implementation time. Therefore, as a matter of configuration, an implementation does not have the ability to suddenly learn how to parse new packet types. However, an implementation may be written such that the software knows where the demultiplexing field is for a particular protocol, and can be written in such a way that the decoding of the next layer up is table-driven. This works when the code has been written to accomodate it and can be extended no more than one level higher. This extensibility is called "limited extensibility" to highlight these limitations. However, this can be a very useful tool. For example, suppose that an implementation has C code that understands how to decode IP packets on any of several ethernet encapsulations, and also knows how to interpret the IP protocol field to recognize UDP packets and how to decode the UDP port number fields. That implementation may be table- driven so that among the many different UDP port numbers possible, it is configured to recognize 161 as SNMP, port 53 as DNS, and port 69 as TFTP. The limited extensibility of the protocol directory table would allow an SNMP operation to create an entry that would create an additional table mapping for UDP that would recognize UDP port 123 as NTP and begin counting such packets. This limited extensibility is an option that an implementation can choose to allow or disallow for any protocol that has child protocols. 5.3. Errors in packets Packets with link-level errors are not counted anywhere in this MIB because most variables in this MIB requires the decoding of the contents of the packet, which is meaningless if there is a link-level error. Packets in which protocol errors are detected are counted for all protocols below the layer in which the error was encountered. The implication of this is that packets in which errors are detected at
the network-layer are not counted anywhere in this MIB, while packets with errors detected at the transport layer may have network-layer statistics counted.