For the campus network Authentication & Billing system, the real dangerous time period is not day but every night 19:30.–23:30。
During this time period, the system faces several superimposed variables:
-
A large number of users are online at the same time.
-
Expiration, renewal, arrears intense trigger
-
Terminals are frequently lined up and down
-
The wireless environment itself is unstable.
If there were minor problems with the design of the costing system, only one result would be:Focus on the net.。
The concentration of the network is almost the most undesired scene in the university information sector.
What do you mean, "one"?“Focused network break risk”It's not a simple line.
In the campus network Authentication & Billing system, so-called centralized disconnects are not usually broken links but rather:
-
A lot of users get used to it in very short periods.Uniform trigger for fee-counting rules
-
Online sessions are recycled in bulk
-
Authentication retriggers, avalanche effects.
From the user side, yes.“The whole dorm's out of network.”，
From the side of the system, it's...The billing engine did the same thing at the same time.。
The essence of this kind of problem is thatTime and state design deficiencies in the billing system。
II. Centralized network breakout, often from 3 common design errors
1 x due time designed as“Point-to-point.”
Many early or low-mature campus networks NATSHEL_AUTH_BILLING systems have a customary fee cycle set as follows:
-
Early at 00:00
-
One day 23:59 due
-
Unique time point batch lapse
In the accommodation environment, this design is equivalent to the active creation of risks.
Once the due time is highly concentrated, the billing system will do hundreds of online sessions in the same second.“Permissions Retrieval”Operation, directly triggers a centralized break.
2 ⁇ Cost state change = forced down
Another typical question is:
Change the billing state to simply equal“Downline User”。
For example:
-
Package expires → Force Disconnect
-
Arrears → Kick off the line now.
-
Package changes → Rebuild Session
This treatment is extremely dangerous during the peak period of accommodation.
A truly mature campus network, NATSHEL_AUTH_BILLING system, no longer in use“All-round kick.”, which deals with changes in the billing status.
3 x 3 Billing Engines do not synchronize with online sessions
If between the billing system and online sessions:
-
Status does not sync
-
Heart beats strong.
-
The session recycling mechanism is rough
So, at peaks, any slight shaking will be magnified.“Systemic issues”。
III. How to dismantle mature billing systems“Focused network break risk”
In the real college project, it's not a problem to avoid concentrating on the network.“Harder Servers”- Oh, shit.Design Selection。
1 Disperse at due time, not uniform
The mature campus network Authentication & Billing system would do one thing at the billing engine level:
Disperse the due date.
For example:
-
Same user, internal maturity time is distributed within a time window
-
For the user, it's still...“One month.”
-
Not for the system.“The same second lapses.”
The result of this is:
-
Expiration event is dispersed naturally
-
Time point for batch recovery sessions not available
2 Episode not equal to a break in the web, but“Policy Switch”
During the peak period of the dormitory, the real stabilization is:
-
Due → Switching to the speed/ permission policy
-
Continuation → Revert the strategy immediately
-
Do not force the break-off of online sessions
The billing engine controls thatNetwork Permissions StatusNot“Online or not”。
The NATSHEL_BILLING system of the NATSHEL_BRAND campus, which is widely used in the accommodation landscape, has significantly reduced the probability of late peak complaints.
3 Schematicization of billing events, not blockage
In a high-separation accommodation environment, the billing system if adopted:
-
Synchronize writing of accounts
-
Sync recovery session
-
Synchronize distribution policy
The peak period will certainly be magnified.
A fee-for-engine for mature systems, usually using:
-
Astrological Incidents queue
-
Online Session Cache
-
The policy delay is effective without interruption of connection
For users, there is little perception but a qualitative improvement in system stability.
4 Sessions“Moderate recycling”♪ Not forced to empty
In the context of the true university project, dormitories are frequently used:
-
Sleep.
-
Lock screen
-
WiFi Wandering
The system must have:
-
Aging the session mechanism
-
Frees the stiff session automatically
-
The abnormal disconnect will not be terminated immediately.
Otherwise, peak-time speech is inflated and ultimately drags down the system.
IV. Anonymous cases: behaviour of the system during the peak period of real accommodation
Below are the anonymous university projects, which logically operate in a true way.
Background
-
School students, about 30,000.
-
5 dormitories
-
Night Summit Online Rate > 70%
-
Billing method: monthly package + terminal limit
Actual performance
-
The package is not forced offline.
-
Auto-entry speed limit policy after due
-
The late peak cost households,–Recovery rate in 5 seconds
-
No focus network breakout
-
The number of complaints was significantly lower than the pre-reform system
The point is not that“Complex rules”And it's...The understanding of peak behavior in billing engines。
V. Why the highs of dormitories are the best test for the maturity of the system
In the campus network Authentication & Billing, the peak period of accommodation is the most real and cruel test environment:
-
No ideal network.
-
No regular behavior.
-
No one's working to intervene in space.
The system of billing that can accommodate the late peaks of accommodation is called a truly long-term system.
In the context of the 21-year college project, it is in this environment that the synergy between the engine and the certification system is constantly being grinded to create a network of campus products that are now stable, cost-effective and suitable for deployment on a scale.