Nectar Sizing Guide

Nectar software products can be deployed using physical servers or virtual appliances. Refer to the corresponding Nectar Installation Guide for the server and product to get more information.

This sizing guide uses the following terms to quantify capacity.

• Session: The VoIP control signaling for a VoIP call may traverse several VoIP call servers. If more than one call server anchors the control signaling for a call, then the call will have multiple call legs. A session is a call leg.

• Sessions per second: This is the maximum number of new sessions that a server can process per second. A new session is identified by the initial control signaling message; e.g. an INVITE for SIP.

• Concurrent Sessions: This is the maximum number of simultaneously active sessions supported by a server.

• RTP streams: A connected VoIP audio call leg will typically have two unidirectional RTP streams. One RTP stream carries the media from the calling party to the called party and the other stream carries the media from the called party to the calling party. Sizing is based on unidirectional streams and an audio session should be assumed to have two streams for sizing purposes. Video sessions with audio should be assumed to have four streams; one audio stream and one video stream in each direction.

• Concurrent RTP streams: This is the total number of simultaneous unidirectional RTP streams a Nectar Diagnostics server can monitor. In the case of a Perspective Agent, this is the total number of simultaneous unidirectional RTP streams that can be generated.

• Perspective Session: A perspective session started on the controller will result in two unidirectional RTP streams involving the two Perspective Agents used. Each Agent sends an RTP stream to the other Agent.

• RTCP streams: Each unidirectional RTP stream may have an associated control (RTCP) stream which may contain QoE metrics measured by the end point receiving the stream.

By default, a Perspective Agent is restricted to 50 concurrent RTP streams. This default can be overridden by modifying the local configuration file. The Perspective Agent can synthesize sending various audio codecs and the H.264 video codec. Since Perspective places a load on the traversed network, it is important to be very conservative with regard to the number of Perspective Sessions started when the network capacity is unknown.

The highest bandwidth usage for the audio codecs is 64 Kbytes/second. This bandwidth only includes the RTP payload bandwidth. Synthesis for H.264 is done for the following average bandwidths: 128Kbps, 384Kbps, 512Kbps, 1Mbps and 2Mbps. These bandwidths can be misleading for two reasons: the bandwidth only includes the RTP payload and the bandwidths are the average over a one second interval. Even though the video frame rate per second is constant, the size of the frames can vary dramatically and therefore, the peak bandwidth usage will be a multiple, e.g. 2 to 3 times, the average bandwidth. To account for the entire packet bandwidth usage, 5% can be added to the payload bandwidth.

The sizing information in the table below represents the maximum achieved when the NUCs were connected via a single Gigabit switch. The NUCs were installed with Centos.

When Nectar Diagnostics is deployed to monitor VoIP Signaling and/or RTP streams, an additional hardware Network Interface Card (NIC) may be required. The NIC can be a Chelsio T520-BT which has 2 x 1Gb copper interfaces, a Chelsio T520-CR which has 2 x 10Gb optical interfaces or an Intel I350-T4 which has 4 x 1Gb copper interfaces. Specific sizing is provided for each. The Intel I350-T4 does not have hardware filtering and therefore should only be used when messages sent to the Diagnostics’ server are relevant to its function.

When the T520-CR (10G) is deployed with Nectar Diagnostics, the total packet rate transmitted on an interface to the UCD server should not exceed the following;

• 4 interfaces: 1.5G per interface

• 2 interfaces: 2.5G per interface

• 1 interface: 3.2G

Nectar Diagnostics UCD-P and UCD-A functions can be deployed on the same server referred to as a UCD-PA. The collocation of these two functions will diminish the UCD-P capacity since the UCD-A processing is performed at a higher priority. UCD-P session sizing is based on the number of sessions established per second. UCD-A RTP streams are sized by the number of concurrent RTP streams active.

Nectar Diagnostics deploys a UCD-M function. The server resources required by this function is multi-faceted and therefore sizing is dependent on multiple factors in the deployment. It is possible to deploy the UCD-M function in combination with the UCD-P or UCD-PA functions. The UCD-M sizing chart classifies the UCD-M as small, medium and large. If any constraint is exceeded for a given size, the sizing must be adjusted upward. If the UCD-M is deployed with the synchronization feature, it is recommended to always use a standalone system.

See the corresponding Nectar Diagnostic’s Installation Guide for more information about the server and server installation.

If Nectar Diagnostics is deployed on a physical server, Nectar Foundation can be deployed on a guest virtual machine. In this scenario, the kernel/operating system used for Nectar Diagnostics also hosts the Nectar Foundation VM. Host CPU and memory resources are allocated to the virtual machine. The percentage of host resources allocated for the VM determine the capacity of Nectar Foundation and the reciprocal decrease in capacity for Nectar Diagnostics.