When WiFi's online, the work of the carrier is not"Everything's fine."But instead, it's about finding things that are quietly accumulating risks in a bunch of seemingly calm daily data. Most times the system doesn't crash suddenly, but it just gets a little worse.——The success rate of user certification has been reduced from 99.5 per cent to 98 per cent, and is not yet to the point of complaint, but if this trend is not noticed, it may fall to 90 per cent in two weeks.
Distribution analysis of the certification success rate and causes of failure
The certification success rate is the most basic health indicator of WiFi certification system, but it is not enough to look at a percentage. The key is to open up the failure: whether the failure ratio is higher due to password errors or because of time spent, or if the server is unavailable?
If the password error is high, then the approximate rate is a problem on the side of the user.——For example, there are lot errors in the code for accounts sent through a given channel. If it is time-consuming, check whether the authentication server and Radius have been responding longer or whether the network links are shaking more. If the server is not accessible, then the alarm is analysed instead of the transporter.
The transport team should establish a daily inspection mechanism to look at the 24-hour certification success curve and the distribution of reasons for failure every morning.——Like 98% to 100% floating up and down.——There is no need to intervene. If we go down for three days, the screening process must be initiated and not wait until users complain.
Ongoing monitoring of loading times on Portal page
Portal Page Load Slow"Hot water boils frogs.". The load may take only a second to get on the line, and the carrier feels fast and nobody cares about it. Then as network traffic increases, the server loads increase, and load time slowly turns into three seconds, five seconds.——By the time users started complaining, it had actually deteriorated for a long time.
The carrier end should be continuously monitored for loading times on Portal pages. There is no need to deploy an external monitoring node. WiFi certification system itself usually has logs that can extract HTTP response times. What the operator needs is a regular number of days of data, and a calculation of P50, P95, P99 distribution of loading time on Portal pages, whether median and long tails are increasing. If so, it will check if the co-processing capacity of Portal servers has become bottlenecked or if there are delays in network links.
Time distribution and trend predictions of online users
WiFi online users are not static, they fluctuate with the seasons, interest rates, and even weather. School scenes, school year openings, drops in cold summer holidays, and peaks in tourist holidays. What is required is not passive acceptance of these fluctuations, but rather a model for peak prediction based on historical data.——Even if it's just a simple trend judgment.——The simultaneous capacity of the system is confirmed sufficiently in advance of peaks.
There is no need for operational complexity: pull the online user curves from the past six months, indicate the date and cause of peaks (opening, vacations, conferences), then make a capacity assessment two weeks before the next round of the same time nodes arrive. These include DHCP address pools, NAT table capacity, Radius distribution count, Portal server load, each confirming any bottlenecks risk.
Impact of changes in distribution of end type on network strategy
The terminal type was relatively simple a few years ago in WiFi: mainly mobile phones and notebooks. Now the IOT equipment is rising rapidly.——Smart sound boxes, television boxes, surveillance cameras, smart door locks, sensors.——While the individual flow of these equipment is small, it is numerous and connected over time and many do not support complex certification agreements.
The WiFi certification system's wi-vi team should regularly draw on the type distribution data of online terminals.——Like a five percent rise to 20% in three months.——That would have to be considered: Is there a rule that separates IOT equipment from VLAN? Does the MAC certification extend? Will the excessive online time of these IOT devices affect the distribution of DHCP addresses?
Automation and retention inspection of Log Audit
The WiFi certification system logs are not only for fault check, but also for compliance. Public security authorities have a clear requirement to keep the public WiFi web logs, usually for six months.
The log itself also needs automated quality checks. For example: Some time periods have suddenly been zero; this may be that the collection component has been hung up rather than actually no one is online. Are there fields where data formats have changed after a version of the upgrade? This leads to the failure of subsequent data analysis tools. These inspections can be written as simple automated scripts, cycle runs, instead of relying on manual random searches by carriers.
The WiFi certification system is a routine operating device."Find out what's wrong in peace."Most times the system is working, but if the crew only investigates a complaint after it has been made, many minor problems accumulate into a major malfunction in silence."Passive fire."The state of the state becomes"Active care"Finally, there is an operational recommendation: make these monitoring and control indicators a routine check list, which takes 10 minutes to clear key data every day. A simple form is enough without a complex screen or automated report.——Whether there are any abnormal success rates today, whether the number of online users is normal range, whether Portal's loading time is high, and how much log storage space remains. Scrubbing every day, getting used to it, most problems can be detected and intervened at an early stage, and they simply don't grow to the point where they need to climb up in the middle of the night.