Networks are increasingly becoming complex and distributed. As a result, problems like hardware failures, performance degradation, resource allocation, bandwidth monitoring and assignment and service provisioning are harder to solve and become real challenges for Network Operation Centers (NOC). The NOC is in constant search for efficient integrated network management systems required to monitor, interpret, and control the behavior and performance of the network, its hardware devices and software resources.
This article discusses the three network management architectures in use today, so by reading it you will be able to understand the differences between each architecture, and by reading the pros and cons of each architecture you will be able to select the right solution, especially when managing a distributed network.
The three network management architectures are:
- Centralized Network Management
- Distributed Network Management
- Hierarchical Network Management
Keep reading to learn more.
Centralized Network Management Architecture
With a centralized architecture a single management system installation monitors the whole network. This installation may consist of one or more servers due to hardware limitations. If more than one server is used, it is considered as a centralized architecture when all servers are located at the same NOC.
As depicted in the above example diagram, with the centralized network management architecture, a distributed network that spans multiple geographical regions is managed from a single Network Operations Center (NOC). NOC operators from each of the regions use clients to remote connect to the centralized management servers located in another region.
With this architecture, the network management system consists of:
- One or more NMS server(s), located in a single NOC. The servers manage the network of the three regions, meaning that they will require Data Communications Network infrastructure (DCN) i.e., routers, switches and Ethernet connectivity from the single NOC they are installed to the remote regions where the network devices are installed to be able to establish management communication.
- One or more clients located at all regions. These clients communicate remotely over DCN to the central server.
Distributed Network Management Architecture
With a distributed architecture, multiple installations of management systems are used to monitor the whole network. Each management system is installed at a NOC that is responsible to monitor a geographical or administration region / domain, i.e. it is a Domain Manager.
As depicted in the above example diagram, with the distributed network management architecture, a distributed network that spans three regions, is managed from three NMS servers each one located at a regional Network Operating Center (NOC). NOC operators of a region use clients to locally connect to their server that manages the part of the network installed at their region.
With this architecture, the network management system consists of:
- Three NMS servers, each located in a separate NOC. Each server manages the (sub) network of the region it belongs, meaning that it will NOT require Data Communications Network infrastructure (DCN) i.e., routers, switches and Ethernet connectivity, from the NOC region to remote regions.
- Clients located at all regions. These communicate locally to their server.
Hierarchical Network Management Architecture
With a hierarchical architecture, multiple installations of management systems are used to monitor the whole network. Each management system is installed at a NOC that is responsible to monitor a geographical or administration region / domain, i.e. it is a Domain Manager. So far this is exactly the same as the distributed architecture, except that the hierarchical architecture adds an additional layer, the Manager of Managers (MoM). This Manager of Managers sits at a higher level and requests information from domain managers. There is no communication between domain managers, information flow follows the hierarchy. The hierarchy can be further expanded by adding additional layers of MoMs and therefore is quite scalable.
Architecture Selection Criteria
Application availability – The NMS clients need to be able to connect to the server in order to allow the user to perform network management functions. Clients connecting remotely to servers may fail to work due to DCN unavailability or instability affecting the user experience.
Hardware Failures – A server or router hardware failure may affect the availability of management, at least till switchover to a redundant hardware is performed. In case the same hardware is used for the network management of multiple macro-regions, a single point of failure is created. Distributed network architectures are more tolerant of network and hardware failures as they avoid single-point-of-failures.
Handling Multiple Time-zones – When operating a large network that spans multiple time-zones, data integrity, data collection accuracy and network time synchronization must be addressed. In such cases, the NMS solution must be capable of handling multiple time-zones, resulting to a dedicated server requirement for every different time-zone.
Maintenance scheduling – When server maintenance is required, local sites must be able to maintain their own schedule. Centralized NMS architectures do not provide such flexibility.
Single Sign-on – Big organizations need to control user authentication and maintain a common database of users for all their IT systems, using also if possible a single password. Geographically distributed organizations prefer distributed users database, handled at each Network Operation Center. The NMS system must be able to connect to such locally maintained user databases (via TACACS, LDAP, RADIUS), which may be a problem for a centralized NMS system.
DCN and Bandwidth expansions – Servers and software at each local site can be appropriately sized to meet the specific needs of each site.
Comparing the architectures
|Reliability||Depends on DCN, due to connectivity between regions.||Very Reliable, no single point of failure. Regions are independent.||Very Reliable and scalable. Regions are independent but also centrally monitored.|
|Scalability||Depends on DCN: as the network expands the DCN bandwidth between regions increases||Very Good, just add servers to increase the managed network size||Excellent, add servers to increase the managed network size and Manager of Managers for control|
|Flexibility||Limited, only one time-zone is possible per server, regions must follow the central region maintenance schedule||Very Flexible, can handle different time-zones; each region can schedule independent maintenance actions||Very Flexible, can handle different time-zones; each region can schedule independent maintenance actions|
|User Experience||Depends on DCN , as DCN performance affects clients remote access to servers||Very Good, local access to servers ensures UI speed and responsiveness||Excellent, local access to servers ensures User Interface speed and responsiveness. MoM gives end-to-end visibility|