Big Video, generally speaking, is an end-to-end service based on the bearer network. It spans the wired and wireless networks, requiring constant high bandwidth, low latency, and low packet loss rate. Therefore, Big Video maintenance needs to start from video service perception and streamline the full process of traditional network maintenance to improve maintenance efficiency and create more value.
Although much more complicated than traditional service maintenance, Big Video maintenance is still carried out based on three basic factors: scenario, process, and standard. Specifically, maintenance scenarios include system commissioning, daily monitoring, fault handling, customer assurance, health check, resource management, and upgrade & switchover. Maintenance processes are used to ensure effective collaboration between different maintenance departments. Maintenance standards are summarized from abundant maintenance practices to ensure high efficiency and reliable operation of the maintenance system.
As for maintenance scenarios, there is no essential difference between the Big Video service and traditional services. Nevertheless, the Big Video service, which spans the entire end-to-end networks, requires collaboration of service maintenance departments, network maintenance centers, local network maintenance centers, customer service centers, community managers, and even service operation departments. Traditional maintenance processes and standards cannot effectively support such a maintenance system.
How to provide more efficient maintenance processes and standards? What we need is data. Only by comprehensive analysis of all kinds of data across all networks can we find out maintenance data correlation, establish a new maintenance KPI system, and help maintenance departments optimize maintenance processes and standards.
ZTE's Vinsight is a Big Video maintenance analysis platform that caters for all maintenance scenarios. The platform collects data from all networks and all maintenance nodes for big data analysis to find out relevance between maintenance indicators and provide support for maintenance process optimization and standard improvement.
Extensive KPI analysis
Vinsight monitors data collected from a wide range of sources, mostly from STB probes. Through multidimensional analysis of KPIs collected by geographically distributed STB probes, Vinsight enables effective evaluation and special optimization of the network-wide video quality in all aspects. It supports KPI analysis by service bearing mode, service type, terminal model, and terminal access mode, and provides statistics of low quality KPIs by channel and by node.
Through multidimensional analysis and comparison, it is easy to find out the cause for quality reduction. Extensive data analysis is applicable to most unified monitoring scenarios.
Intensive KPI analysis
For a single user, Vinsight can also collect rich KPIs, which spans from TCP/IP protocol stack layer 3 to layer 7, indicating network and service health state at all levels. Such vertical in-depth data analysis can also provide effective data support for daily maintenance.
- The device-related KPIs indicate device health, such as CPU/memory usage and WiFi signal intensity.
- Service KPIs reflect video service quality and user experience, such as number of video freezing times and duration.
- Video service KPIs show video service support capabilities, such as video segment download quality and segment size, TS download speed, HTTP link setup response, and HTTP error codes.
- Network KPIs show network support capabilities, such as TCP link setup time, number of TCP retransmissions, and number of out-of-order TCP segments.
Different combinations of abnormal KPIs from these four aspects give different root causes for the same problem. For example, a problem of video freezing needs to be solved. If the HTTP response is slow and error codes increase, but the TCP response is normal, we know that the CDN server is not capable enough of processing plenty of user requests. If the HTTP response is normal, but TCP retransmission and out-of-order occur frequently, we know that the network packet loss rate is high. If TCP retransmission and out-of-order slightly increase, but the TS download is slow, we know that the network end-to-end bandwidth is insufficient to support high-bandwidth video services. If the terminal’s CPU usage is high and the TCP window size becomes smaller, we know that the terminal does not have adequate performance to promptly resolve or play video streams.
End-to-end KPI system based on professional networks
Figure 1 Big Video End-to-End KPI System
Now, we have the ability of analyzing all KPIs vertically and horizontally, so that we can establish an end-to-end KPI system to make systematic monitoring and analysis in a matrix.
Horizontally, the matrix has three parts: user experience V-QoE, network health, and platform health (see Figure 1).
Each part can be further divided. For example, user experience V-QoE can be classified into video client indicators and home network indicators; network health can be evaluated by access network, transmission network, aggregation network, and backbone network; platform health can be evaluated by CDN network, service platform, and content source.
Each part of the KPI system reflects the major factors that affect end user experience on the corresponding network. In particular, the user-side KPIs are most comprehensive to reflect all-round user experience. Network KPIs mainly reflect networks’ bearing capability of Big Video services. Service KPIs mainly reflect service systems’ capabilities of serving Big Video services.
O&M process and standard optimization by Vinsight
Vinsight provides rich methods of data analysis based on rich experience. It also provides some functions for process optimization.
- Fault delimitation
Maintenance engineers can initially delimit a fault after entering relevant user’s account, fault occurrence time, and symptom.
- Fault knowledge tree
The fault knowledge tree is composed of process-based data check points which are summarized from maintenance experience and data analysis results. The system can automatically analyze fault causes against the fault knowledge tree.
According to the maintenance data provided by Vinsight, maintenance departments can optimize Big Video service processes and standards, for example, the multi-level (province-city-street) maintenance collaboration process, based on which, the user complaint handling process can also be optimized.
In conclusion, Vinsight, the Big Video maintenance analysis platform, performs big data analysis based on data in network positions and on maintenance nodes and uses automatic and intelligent maintenance methods to provide data basis for Big Video maintenance processes and standards.
Zhang Yuan is Big Video Smart Maintenance Product Manager at ZTE Corporation
- ZTE and Sany deepen 5G cooperation to promote the digital transformation of intelligent manufacturing
- ZTE and China Mobile launch a marine broadband satellite solution creating a new era of marine communication
- ZTE common core enables intelligent and simple 5G construction
- ZTE & China Mobile's joint-effort in realizing industry's 1st typical-service-based NB-IoT massive-capacity evaluation with NMVP
- Huawei introduces autonomous driving core network