Deterioration of the quality of Internet access [Документация VAS Experts]

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revisionPrevious revision
Next revision
Previous revision
en:dpi:qoe_analytics:cases:network_health:full_netflow_analytics [2025/10/31 11:54] elena.krasnobryzhen:dpi:qoe_analytics:cases:network_health:full_netflow_analytics [2025/12/11 10:15] (current) – created elena.krasnobryzh
Line 1: Line 1:
 {{indexmenu_n>1}} {{indexmenu_n>1}}
-====== Detecting degradation in internet access quality ====== +====== Deterioration of the quality of Internet access ====== 
-<note important>[[en:dpi:dpi_options:opt_statistics:statistics_ipfix|DPI exports information about all client sessions in IPFIX (NetFlow v10) format.]]</note>+<note important>[[en:dpi:dpi_options:opt_statistics:statistics_ipfix|DPI exports information about all client sessions in IPFIX format (NetFlow v10)]].</note>
  
-The DPI exports [[en:dpi:qoe_analytics:qoe_gui:reports:paramdescript|information about delays]] between the client and the DPI, and between the DPI and the host during TCP session establishment — [[https://en.wikipedia.org/wiki/Round-trip_delay|RTT]]. +The simplest and most studied DPI metric for communication quality testing is [[https://en.wikipedia.org/wiki/Round-trip_delay|RTT]]. DPI allows you to measure RTT and divide it by directions (to the subscriber and from the subscriberand specify by protocols and devices if necessary. A large delay for a long time "from the subscriber" most likely indicates that the subscriber is experiencing difficulties with access to online services - games, video, communication. As rule, the delay occurs due to the subscriber’s WiFi network, but it can also talk about overloading network nodes
-The statistics record latency for each protocol, linked to the UserAgent (taken from ClickStream), which allows tracking the performance of specific device.\\ + 
-Required steps to investigate+Action
-  - Go to QoE Analytics → Subscribers → Netflow+  - Go to the section QoE Analytics → Subscribers → Netflix
   - Create a filter where:   - Create a filter where:
-  * It is recommended to limit the search to http/https protocols to exclude the behavior of other protocols during TCP session setup+    * It is proposed to limit the search by http / https to screen out possible features of other protocols when installing a TCP connection
-  Specify average speed to select subscribers who actively use the internet. +    indicate the average speed to make a sample of subscribers actively using the Internet 
-  Specify the lower RTT threshold from the client side+    specify the lower threshold RTT from the client 
 + 
 +{{:en:dpi:qoe_analytics:cases:queries_from_database:problematic_subscribers_3.png?nolink&700|}} 
 + 
 +It is recommended to work with subscribers who have an average (and even better median) RTT of more than 100 in 24 hours. This can be done by: 
 +  - Upload data daily to your database via the API or directly from Click House QoE. 
 +  - Make a trigger in a graphical interface with specified parameters.
  
-{{ :en:dpi:qoe_analytics:cases:network_health:full_netflow_analytics:rtt_filter.png?600 |}}+When processing, it is important to consider the following factors: 
 +  - It is worth checking the geographical distribution of the “problem” unloading. “Accuracy” usually occurs due to network congestion or an old switch, the problem can be fixed on the operator’s side. It happens that a “bad” area is connected by outdated technology (DSL or radio access), then a bad RTT is no longer a bug, but a feature. 
 +  - The subscriber may not experience discomfort due to the delay (and speaks directly about it if asked by phone), for example in cases of: 
 +    * IoT devices are working, they may have a weak signal 
 +    * Works on outdated equipment and does not expect miracles 
 +    * Do not use online services 
 +  - Problems in Wi-Fi networks can occur due to: 
 +    * Noisy Wi-Fi range 2.4 in an apartment building 
 +    * router 
 +    * complex topology or large size of the apartment / house 
 +  - Sometimes it is impossible to determine the host where the subscriber has contacted with a high RTT.\\ {{:en:dpi:qoe_analytics:cases:queries_from_database:problematic_subscribers_4.png?nolink&1100|}}\\ This can be for three reasons: 
 +    * IP 
 +    * http protocol 
 +    * No SNI. 
 +  - In addition to RTT, you can use the % retransmit metric when searching for problem subscribers. If for a long time (>30 minutes) there is a significant excess of the background value for retransmits (usually 4-5%), this is a sign of degradation of the subscriber’s service. It is not recommended to use retransmitts without RTT, there is a high probability of false positives.
  
-Interpreting the resulting statistics+{{:en:dpi:qoe_analytics:cases:queries_from_database:problematic_subscribers_5.png?nolink&1100|}}
  
-{{ :en:dpi:qoe_analytics:cases:network_health:full_netflow_analytics:rtt_result.png?600 |}}+Example of a request to the database: 
 +{{:dpi:qoe_analytics:cases:queries_from_database:subs_bad_rtt_sample.sh |Download script}}
  
-  * The filter identified 25 potential clients who may have access quality issues. +Parameters in the script:\\ 
-  * Detailed latency information for these clients can be viewed in the **"Details"** window. +''format='CSV'' is the output format. By defaultCSV. Possible formatshttps://clickhouse.com/docs/en/interfaces/formats/\\ 
-  * Using the megaphone iconthey can be added to a __[[en:dpi:dpi_components:dpiui:user_guide:ssg_control_section:ad_campaign_management|marketing campaign]] and notified or surveyed through the browser regarding service satisfaction__+''periodSecs=24*3600'' is the period in seconds. The default is 24 hours.\\ 
-  * The report can be exported in a convenient format.+''rttMore=100'' is the value of RTT. By default, 100.