Network bandwidth

Bandwidth requirements

In order to effectively manage the network, NFM-P must have access to sufficient bandwidth between the NFM-P server(s), NFM-P auxiliary(s) and the network elements.

This bandwidth will be used to carry the management traffic between NFM-P and the network element. The following table describes the bandwidth requirements for a particular network element.

Table 4-11: NFM-P server to network bandwidth requirements

Network element example		Bandwidth requirement from NFM-P server(s) to the network element
7950 XRS		2-4 Mbps
7750 SR-12E (fully loaded)		2 Mbps
7750 SR-12 (fully loaded)		2 Mbps
7750 SR-2s		2 Mbps
7750 SR-a4		1 Mbps
7750 SR-c12 (fully loaded)		600 kbps
7450 ESS-7 (fully loaded)		1 Mbps
7450 ESS-1		200 kbps
7705 SAR (fully loaded)		200 kbps – 400 kbps
7250 IXR-6 / 7250 IXR-R4 / 7250 IXR-R6 / 7250 IXR-R6d / 7250 IXR-R6dl / 7250 IXR-x		800 kbps – 1000 kbps
7250 IXR-e / 7250 IXR-e2		300 kbps
7210 SAS-E, 7210 SAS-M, 7210 SAS-K		200-300 kbps
7210 SAS-D, 7210 SAS-X, 7210 SAS-T, 7210 SAS-R, 7210 SAS-Mxp, 7210 SAS-Sx		500-600 kbps
7701 CPAA / vCPAA		250 kbps
9500 MPR / Wavence SM		200 kbps
OmniSwitch 6250, 6350 6400, 6450, 6465, 6560, 6850, 6855, 6865, 9000 Series		300 kbps
OmniSwitch 6860, 6860E, 6860N, 6900, 10K		400 kbps
1830 VWM OSU		400 kbps

Details on the bandwidth requirements

The recommended bandwidth described above is a conservative figure that is meant to ensure that the performance of NFM-P and its ability to manage successfully each network element will not be affected by unusual network conditions.

Specifically, the bandwidth recommendation ensures that NFM-P can fully discover (or resynchronize) all of the objects contained in the network element, within a reasonable amount of time, varying heavily based upon the specific network element type and configuration.

The following are the main operations that result in significant amounts of information being exchanged between NFM-P and the network elements. These factors are therefore the principal contributors to the bandwidth requirements.

Network element discovery: Upon first discovery of the network element, a significant amount of data is exchanged between NFM-P and the network element.
SNMP traps: SNMP traps do not result directly in significant data being sent from the network element to the NFM-P. Several of the SNMP traps however do not contain all of the information required for NFM-P to completely represent the new status of the network element. As a result, NFM-P will subsequently perform a poll of a certain number of the SNMP MIBs to obtain the required information from the network element. Consequently, SNMP traps do result in a certain quantity of data and therefore cause bandwidth utilization. The exact quantity of bandwidth utilized will vary based on the number and the type of trap that is sent from the network element. In the worst case however, this bandwidth utilization will be less than that utilized during a network element discovery.
SNMP polling: It is possible to configure NFM-P to poll the SNMP MIBs on the network elements at various intervals. By default, NFM-P will perform a complete poll of the SNMP MIBs every 24 hours on non-SR–OS based network elements. During the polling cycle, the amount of data transferred between NFM-P and the network element is equivalent to the amount of data transferred during the network element discovery.
Statistics collection: It is possible to configure NFM-P to poll the SNMP MIBs on the network elements that contain performance statistics information. During the polling cycle, the amount of data transferred between NFM-P and the network element is less than the amount of data transferred during the network element discovery. With the configuration of an NFM-P auxiliary statistics collector, the communication from and to the network elements will be distributed between the NFM-P server and an NFM-P auxiliary statistics collector.
Network element backup: It is possible to configure NFM-P to request a backup of the network element at specified interval. During the NE backup cycle, the amount of data transferred between NFM-P and the network element is less than half of the amount of data transferred during the network element discovery.
Provisioning of services and deployment of configuration changes: When network elements are configured or when services are provisioned via the NFM-P GUI or via application using the API, a small quantity of network bandwidth is utilized. The amount of data transferred is significantly less than during the network element discovery.
Initiation and collection of STM tests and their results: When STM tests are initiated, the NFM-P server sends individual requests per elemental test to the network elements. Once the test is complete, the network elements report back using a trap. The NFM-P server then requests the information from the network element, and stores it in the database. This can result in a significant increase in network traffic to the network elements.
Software Downloads: The infrequent downloading of network element software loads is not included in the bandwidth levels stated in Table 4-11, NFM-P server to network bandwidth requirements. Bandwidth requirements will depend upon the size of the network element software load and the desired amount of time to successfully transfer the file to the NE.

For some network elements, management of the NE includes methods other than standard MIB/SNMP management – for example web-based tools. These network elements may require additional bandwidth above the bandwidth levels stated in Table 4-11, NFM-P server to network bandwidth requirements.

Possible consequences of insufficient bandwidth

In situations where there is less than the recommended bandwidth between the NFM-P and the network element, the following are possible consequences:

The length of time required to perform a network element discovery will increase
The length of time required to perform a SNMP poll of the network element will increase
The length of time required to retrieve statistics from the network element will increase
The proportion of SNMP traps that will not reach NFM-P because of congestion will increase. This is significant since NFM-P will detect it has missed traps from the network element and will result in NFM-P performing additional SNMP polling to retrieve the missing information. This will result in additional data being transferred, which will increase the bandwidth requirements, possibly exacerbating the situation.

Determining total bandwidth requirements for NFM-P managed networks

The amount of bandwidth required for each of the network elements should be obtained from Table 4-11, NFM-P server to network bandwidth requirements.

The total amount of bandwidth that is required for NFM-P to manage the complete network will vary based on the topology of the infrastructure that is used to carry the management traffic. From NFM-P’s perspective, there must be sufficient bandwidth (as per Table 4-11, NFM-P server to network bandwidth requirements) between itself and each of the network elements that is under management.

In cases where the management traffic is carried over physical point-to-point links between the NFM-P server and NFM-P auxiliary network and each of the network elements, sufficient bandwidth must be reserved on the physical links. The NFM-P server complex can simultaneously communicate to several NEs for the following functions:

NE discovery, NE resync, resyncing for trap processing
NE backups, NE software downloading, and sending configurations to NEs
Collecting performance statistics
Collecting accounting statistics
Initiating STM tests on NEs
Retrieve STM Test Results - also via (s)FTP
NE reachability checks and NE trap gap checks

Rarely are all of the above performed simultaneously so it is recommended to assume for link aggregation points that NFM-P can communicate with a minimum of 20-30 NEs simultaneously – this can increase to 60-70 NEs on a 16 CPU core NFM-P server station. For Networks of over 1000 NEs or where an NFM-P auxiliary statistics collector is being used, that number should be increased by 20-30 NEs. Higher bandwidth maybe required under special cases where above average data is attempted to be transferred between NFM-P and the network elements. For example, large statistics files, NE backups, or software images.