On Network Management for the Internet of Things

Qinghua Wang, Riku Jäntti, Yusein Ali Aalto University

School of Electrical Engineering Department of Communications and Networking

PO Box 13000, FI-00076 AALTO, Finland {qinghua.wang, riku.jantti, yusein.ali}@aalto.fi

Abstract—This paper studies network management issues for a new communication paradigm – the Internet of Things (IoT). Basic management functionalities and requirements for the IoT management are discussed. A management structure is proposed. In the end, an extension of the SNMP protocol is suggested for the IoT management.

Keywords- network management; Internet of Things; SNMP

I. INTRODUCTION The Internet of Things (IoT) is a new communication

paradigm which extends the traditional human-to-human communication paradigm to the more modern thing-to-thing or thing-to-human communication paradigm. It is an area that the traditional communication technologies including the Internet, cellular networks and satellite networks, and the future communication technologies including wireless sensor networks, machine-to-machine communication, RFID, merge. A characteristic of traditional networks is that there are heavy infrastructures, such as routers, base stations, power lines, etc, while the IoT extends the traditional networks to non-infrastructure communication with simple and easy-to-deploy intelligent devices, such as sensors and RFIDs.

Because the IoT is a new communication paradigm, it also introduces new challenges to network management. The basic principle of network management is shown in Figure 1, where the IoT operation is constantly monitored by the management facilities. If there is a change in the management policies or configurations, this is also reflected in the IoT operation and can be monitored. The network-level management acts according to the real-time network operation status and the changed network management requirements. This is pretty much like a closed-loop control system.

Figure 1: Network management

What makes the IoT management especially challenging includes the scale of the network size, the dynamic network operation status, and the resource constraints. A good network management system for the IoT should be able to react to dynamic network operation changes in real time and to fulfill network resource constraints, such as energy and memories, and to provide a good scalability and interoperability with infrastructure networks.

In the following, an in-depth discussion about the IoT management, its requirements, functionalities, challenges and solutions will be given.

II. IOT IN OUR VISION The Internet of Things (IoT) refers to the technology where

our future Internet is enabled to communicate with “things” (e.g. everyday objects, cars, power meters, air conditioners, etc.). It is a place where different technologies, such as the Internet, cellular networks, wired and wireless sensor and actuator networks, RFID, and machine-to-machine communication, etc., merge. With the architecture of the IoT, many new and exciting applications could be envisioned. The most exciting application scenarios include: a remote healthcare application where doctors can monitor the healthy status of patients remotely via the IoT; a smart grid application where increased use of digital information and controls technology are used to improve reliability, security, and efficiency of the electric grid; an environmental monitoring application where dangerous places such as volcanoes, glacier, and disaster scenes can be monitored; an intelligent traffic application where cars can talk with traffic signs, cellular phones as well as other cars to significantly improve traffic security and optimize traffic flow; etc. A sketch map of the IoT is illustrated in Figure 2. A recent complete survey of different enabling technologies and different visions of the IoT is given by [1]. To enable this kind of intelligent communication and cooperative action among “things” and between “things” and the Internet, these “things” are extended with sensors, RFIDs, actuators, micro processors and communication transceivers which in together turn dead “things” into smart “things”. Ericsson estimates there will be 50 billion interconnected intelligent “things” by 2020, which would amount to around 10 “things” for every human on the earth.

Figure 2: Internet of Things

III. NECESSITY OF THE IOT MANAGEMENT The popularization of the Internet has been accompanied by

the implementation of different kinds of network management protocols. SNMP [2] has been widely used for monitoring fault and performance of network components in the Internet. SYSLOG [3] allows a host to send system log messages across the Internet to an event message collector. IPFIX [4] can be used to perform usage accounting, traffic profiling, anomaly detection and QoS monitoring based on flow-based traffic measurements. NETCONF [5] provides mechanisms to configure network devices. RADIUS [6] and DIAMETER [7] are protocols which can carry authentication, authorization, and configuration information between a network access server and a shared authentication server.

The IoT is expected to expand the future Internet with at least ten times more connected devices. The new network paradigm will not function efficiently if there are no efficient network management mechanisms. In the simplest scenario, there will be a need of management which is comparable to the ones adopted by the current Internet. The interconnection of wireless sensor networks, RFID and machine to machine communication networks with the Internet-like infrastructure networks will require management functionalities being extended to these new paradigms of communication. For example, the collection of energy states will be extremely important for energy-aware operation of energy-constrained sensor networks. Data compression and data fusion will be extremely important due to the large amount of data. Management protocols also need to tackle the challenges of dynamic network topology, prone-to-failure devices, constrained computing and communication resources, heterogeneous device platforms and heterogeneous applications.

IV. KEY MANAGEMENT FUNCTIONALITIES OF THE IOT The OSI network management model developed by ISO

has defined five key network management functions, namely fault management, configuration management, accounting management, performance management and security management. This also applies to the IoT.

1) Fault Management: fault management performs network monitoring and detects abnormal operations. In order to localize a fault, a sequence of diagnostic tests may be carried out. In response to a fault, actions which can bypass or repair the fault will be launched. Usually, fault management also logs error information.

2) Configuration Management: configuration management maintains an accurate inventory of network resources (i.e. hardware, software and communication links). It also updates resource allocation according to changes in service requirements and changes in system states. In order to fulfill these functionalities, it must track all changes made to network resources.

3) Accouting Management: accounting management identifies cost for use of network resources, and gathers usage statistics for users. It provides information for billing and charging individual users or service subscribers.

4) Performance Management: performance management takes care of the monitoring of network resource utilization. It needs to ensure that user service level objectives are met while the network resources are utilized in a cost-efficient way. It usually adopts some network management protocol to regularly gather network performance data such as network response time, packet loss rates, link utilization, etc. It performs performance evaluation and reports that to administrators. It may trigger fault diagnosis and network reconfiguration.

5) Security Management: security management mainly takes care of network access control and information protection. It includes managing network authentication, authorization and auditing such that any user only has access to appropriate network resources. It also regularly gathers security-related information and analyzes them. In response to security threats, it may involve the tasks of configuring network firewalls, intrusion detection systems and security policies. To protect information from being disclosed and modified, cryptography is usually adopted. Algorithms and keys used to encrypt messages are also managed by security management.

V. REQUIREMENTS ON THE IOT MANAGEMENT Considering different factors such as the interoperability

with infrastructure networks, the resource constraints of the IoT, the prone-to-failure characteristic, fast changing topology and other dynamics in the IoT, there are many requirements imposed on the IoT management.

1) Interoperability with Infrastructure Networks: the IoT management should allow an easy interoperation with the infrastructure networks especially the Internet. This means the protocols and interfaces used in the IoT management should also be able to be easily supported by the Internet management. Ideally, Internet management models and protocols should be extended to support the IoT management.

2) Low Overhead: management should cause as little overhead as possible for the underlying networks considering

the resource constraints of the IoT. To fulfill this, unnecessary communication should be avoided. For example, broadcast and multicast messages could be used to replace unicast messages if applicable. History query results could be stored at some proxy points for future queries instead of flooding all query messages into the network. Message headers and data should also be compressed if possible.

3) Self-Organizing: the IoT networks are mostly self-organizing networks by themselves. Therefore, the management for the IoT should also be able to self-organize itself. In the case of node failure, link failure and other uncertainties, the IoT management should be able to automatically detect a change and adapt the management behavior according to the change.

VI. A NEW MANAGEMENT ARCHITECTURE FOR THE IOT As it is said in Sunzi’s Art of War, “Know yourself. Know

your enemy. You will never be defeated”. Efficient management requires the understanding of the managed entities’ status, as well as a set of well defined actions that could be adopted to manage the entities. In the IoT, system operation states change fast. Real-time monitoring demands a set of tools and protocols which could regularly take snapshots of a managed entity’s states and report those to a manager. In the case of emergency, alerts must also be launched. This kind of reporting and alerting tasks could be done ideally by a piece of software called agents which reside inside the managed entities. In order to perform remote management, there is another software tool with user interface which is called manager. The manager could be installed on a management server. The manager receives configuration inputs from an administrator, makes configuration changes, sends commands or queries to the management agents through the help of a management protocol, and receives query results or alerts from the management agents through the help of a management protocol. The manager also maintains a management information database (MIB) about network-wide system states (and another way is to distributedly store MIB information at the places of agents).

What makes IoT different from the Internet is that there are many autonomous device communication networks such as sensor networks, machine-to-machine communication networks and RFID networks. These device communication networks are connected to the Internet or a cellular network via an IoT gateway. What we can utilize this network architecture for our management purpose is that we can have management agents reside inside devices, a local manager installed at the place of an IoT gateway, and a remote manager installed somewhere in the Internet. A sketch map of this IoT management architecture is shown in Figure 3.

This network management architecture allows the interoperation of infrastructure network management facilities with the management facilities residing inside the IoT devices. In this case, a local manager which resides in a gateway could function as a cache for received IoT network status and provides indirect query results after receiving a query request from a remote manager. By avoiding pushing all queries to the

Figure 3: IoT management architecture

IoT devices, the caching of management information at a local manager saves the precious energy resources in the IoT devices. In the case that different management protocols or different versions of a management protocol are used for the IoT devices and for the infrastructure networks, a local manager could function as a proxy by providing translation services between different protocols.

VII. ADAPTION OF SNMP FOR THE IOT MANAGEMENT SNMP [2] is a simple network management protocol which

is widely implemented in the Internet. The simplicity and wide popularity of SNMP decide that it could also be adapted to the management of the IoT. The adoption of SNMP or a modified version of SNMP will allow the interoperation of the management functions between the infrastructure networks and the IoT-specific networks.

1) Standardized SNMP Communication: the objective of SNMP is to provide an efficient and standardized protocol for the communication between the management entities. Therefore, an important question is how to adapt SNMP messages for the communication of the IoT management. Given the structure of IoT management in Figure 3, the essential communication messages between the management entities include: The control and configuration messages sent from a manager to an agent; the regular monitoring messages sent from an agent to a manger; and the alarming messages sent from an agent to a manager. According to [2], SNMP has provided different PDUs (i.e. protocol data units) which could serve the management communication purposes here. For the control and configuration, a SetRequest PDU can be sent from a manager to an agent, and the destined agent could reply with a GetResponse PDU. For the monitoring purpose, a GetRequest PDU could be sent by a manager to one or a group of agents, and the destined agents could reply with a GetResponse PDU. For the alarming messages, a SetRequest PDU could be sent from a manager to an agent to specify the desired alarm types and their thresholds, and a Trap PDU or an InformRequest PDU (which is standardized in SNMP v2 [8]) could be used to send alarm messages when the alarm conditions are fulfilled.

2) Extensions of SNMP Communication: the IoT is characterized by a large number of cheap and simple devices connected to an infrastructure network via a gateway. Since

the large number of cheap and simple devices within an autonomous network may only perform similar tasks (e.g. temperature sensing), it could be efficient by broadcasting or multicasting the same management configuration/control messages to multiple recipients. However, the standard SNMP protocol has not specified any broadcast/multicast PDUs, it has been suggested in [9] that Boradcast/Multicast SNMP messages should be adopted by SNMP for its application over IPv6-enabled low power communications such as the situation in the IoT. Another extension suggested in [9] is the Periodic GetRequest PDU and the Stop PeriodicGet PDU. This extension is vital for the IoT from an energy efficiency point of view. As regular monitoring messages are expected to be sent from devices to a manager, it is desired that this periodic data reporting is only requested once by the manager. In order to cease the process of periodic data reporting, A Stop PeriodicGet PDU should also be defined. Other modifications of SNMP could be the compression of SNMP messages to save the energy consumption. In order to perform compatible operation between the extended SNMP and a standardized SNMP, a proxy should be implemented for protocol translation. Considering the fact that the extended version of SNMP may only be implemented in the IoT-specific networks, this proxy could be implemented as a standard function of IoT gateways. In Figure 3, the local manager could also function as a proxy between the IoT management agents and a remote manager located inside the infrastructure network.

VIII. CONCLUSIONS Internet of Things (IoT) is a new communication paradigm

which has many similarities to the traditional networks, but also many IoT-specific characteristics. In this paper, we present management architecture for the Internet of Things. Important management functions and requirements are discussed. We also suggest the adoption of SNMP in the IoT management.

ACKNOWLEDGMENT This work is done under the framework of the Finnish

Tekes-funded SHOK programme on the Internet of Things (TIVIT Oy).

REFERENCES [1] L. Atzori, A. Iera, G. Morabito, “The Internet of things: a survey”,

Computer Networks, vol. 54, issue 15, 2010, pp. 2787-2805. [2] RFC 1157, A Simple Network Management Protocol (SNMP), May

1990. [3] RFC 5424, The Syslog Protocol, March 2009. [4] RFC 5101, Specification of the IP Flow Information Export (IPFIX)

Protocol for the Exchange of IP Traffic Flow Information, January 2008. [5] RFC 4741, NETCONF Configuration Protocol, December 2006. [6] RFC 2865, Remote Authentication Dial In User Service (RADIUS),

June 2000. [7] RFC 3588, Diameter Base Protocol, September 2003. [8] RFC 3416, Protocol Operations for Version 2 of the Simple Network

Management Protocol (SNMPv2), December 2002. [9] H. Choi, N. Kim, and H. Cha, “6LoWPAN-SNMP: Simple Network

Management Protocol for 6LoWPAN”, 2009 11th IEEE International Conference on High Performance Computing and Communications.