With the advent of the age of big data and artificial intelligence, more and more data are being aggregated. In such an environment, it became a problem how to obtain more effective and pure data and deliver it to the recording and recording server of the call center?

In a centralized big data center, we usually get the raw data of the audio and video recording through the mirror port of the network switch. In order to more effectively manage and distribute the mirrored data, we can use SpanX series DC-TX124x to collect, filter, and distribute data. The deployment of N+N or N+1 audio and video solutions can therefore become flexible.

When artificial intelligence engine needs real-time forwarded data, in order to ensure non-destructive data distribution, we can use the DC-TX4G of SpanX series to increase the processing performance of standard servers and reduce or balance the load on the server CPU.

 

SpanX is designed for high-density and high-reliability VoIP voice and video recording systems. It can intelligently collect, filter, aggregate and distribute data, significantly reduce the loss rate of audio and video, ensure the validity of data, and improve system reliability.

 

Case Introduction

Customer’s enterprise voice communication is based on SIP protocol communication. The IP switch uses a dual-server load balancing mechanism. Some customers’ voice data is connected to the IP switch through the SIP trunk. Some customers’ voice data is directly registered to the IP switch through the SIP phone terminal. There are 2 SIP trunks (redundant, total 4), the number of telephone terminals is approximately 10,000, and the total number of concurrent terminals is approximately 60%.

 

System Structure

Figure-1 Overall Architecture

Devices such as voice switches and recording systems are deployed in the operation center. Enterprise users register to the center through the network or SIP trunk and call the PSTN through the operation center. The voice data between the internal extensions in the enterprise will also pass through the voice switch of the operation center. The IP data nondestructive distributor DC-TX1240, SIP signaling analysis and management server, recording server, database, and Web server all need to access the office LAN.

Figure-2 Recording related device link architecture

Two load-balanced voice switches, PBX-1 and PBX-2, provide two redundant mirrored ports and are connected to the data splitter active and standby machines respectively:

(1) PBX-1 provides a set of redundant mirror ports SP1-1 and SP1-2. The two mirror ports transmit the same call data at the same time. SP1-1 connects to the DC-TX1240 active server and SP1-2 connects to the DC-TX1240 standby server.

(2) PBX-2 provides a set of redundant mirror ports SP2-1 and SP2-2. The two mirror ports transmit the same call data at the same time. SP2-1 connects to the DC-TX1240 active server and SP2-2 connects to the DC-TX1240 standby server.

The IP data nondestructive distributor DC-TX1240 (one host and one standby) splits the mirrored data of the call into SIP signaling data and RTP voice data. The SIP signaling data is sent to the SIP signaling analysis and management server. The RTP voice data is distributed to each recording server according to the port number rules. By splitting and diverting data, the operating pressure of a single server can be reduced.

 

DC-TX1240IP data nondestructive distributor active and standby working mechanism:

(1) The DC-TX1240 active server sets the parameters of the data distribution rule and takes into effect (the parameters of the active and standby devices are the same). The device is in working state and receives and processes the mirrored data from SP1-1 and SP2-1.

(2) The DC-TX1240 standby server sets the parameters of the data distribution rule but does not take into effect (the parameters of the active and standby devices are consistent), and the device is in a standby state. When the management server detects that the DC-TX1240 host is not functioning properly, it sends a valid command to the standby server to start and receive mirrored data from SP1-2 and SP2-2. At this point, the DC-TX1240 standby machine becomes the new active server, and the original DC-TX1240 active server becomes the backup server after being restored and is in a standby state.

SIP signal analysis and management server (1 server and 1 standby) : This device is responsible for parsing the SIP signaling and instructs the data splitter to distribute the RTP voice data to the corresponding recording server according to the protocol, port number, and other rules, and notify the recording. Server starts/stops recording

In addition to analyze SIP signaling, this device is also responsible for monitoring and centrally managing the operational status of the DC-TX1240 data splitter.

 

SIP signaling analysis and management server main and standby working mechanism:

Under normal circumstances, there is pulse detection between the active server and the standby server, and the data is synchronized; when the active server is down, the standby server takes over the work and becomes the new active server; the original active server becomes the new standby server.

  • Recording server (host 3, standby 1)

With N+1 backup, each recording server can carry 500 calls.

DC-TX4G is used on the recording server to receive RTP data stream.

When one of the recording server hosts does not work, the standby machine takes over its work. After taking over, the standby machine becomes the new host, and the original host becomes the standby after recovery.

  • Database and Web Server

Deploy the Oracle Enterprise Edition database. The system adopts dual-system redundant hot standby mode, with one active and one standby server, and maintains data synchronization. Enterprise users can access both the active and standby servers when they query through the Web, thus allows backup while balancing query traffic.

The above solution achieves an N+1 recording solution with high reliability and stability.