The site is often more difficult to handle when a network billing system fails and users are unable to access the Internet than other systems fail in hotels or commercial locations, because users are present and immediately aware. The first complaint received from the front desk usually comes within minutes of the failure.
Many operators realized that they were not carefully designed to accommodate disasters when the system was deployed. The billing server was hung up, no backups; the database was damaged, there was no recent backup; authentication services were interrupted and all users were offline. It took hours or more to restore, during which time it would be possible to open a free network manually or leave users without access to it.
This is a very clear statement.Network billing systemWhat basic requirements are to be met in terms of disaster response and backup, and what a rational implementation strategy should be at different scales of deployment.
Baseline requirements for service availabilityThe cost-resistance system is a starting point for disaster design. Different components of the system require different requirements for usability. Certification services (user networked doorcards) are the most demanding, and the machine can affect all online users for one minute; after billing records services (flow and time measures), short interruptions can be compensated by caches; billing and payment services allow certain delays, and users will not be immediately out of reach because they have been interrupted. Setting differentiated availability targets for different components allows limited resources to be used where critical.
Localization of certification servicesThe core idea is: to deploy a local authentication cache at each site, which normally synchronizes the authentication data from a central server; when network breaks or central servers are unavailable, local equipment takes over certification and uses the cached authentication data to continue processing user connection requests."Central server is disconnected or network is disconnected"The resulting disruption of local services is a common response to disasters in distributed deployment sites (e.g. hotel chains).
Data backup is the basic work in disaster-prone design. The following parameters need to be specified: frequency of backup (a few times a day? Real real-time incremental or regular full volume), length of time for backup retention (how many days of historical backup are kept), location of backup storage (whether physical isolation is available from the main server and whether offsite backup is available), backup recovery exercises (when do you conduct a recovery test to confirm that backup data is actually available). Putting backup data on the same machine as the main server is a very common but very dangerous error——When hard disks fail, backup and raw data are lost simultaneously.
Automation of failure switchPrimary disaster-response programmes require manual diagnosis of malfunctions, manual switching to backup systems, which can take about 10 minutes; mature programmes have automatic failure detection (heart beat monitoring + threshold warning) and automated switch mechanisms that can be reduced to less than one or two minutes. Automatic switching is more complex in achieving the transition, but at critical times when a failure is made it can significantly reduce the number of hours that users perceive as service breaks.
In selecting a disaster-relief programme, one practical problem is the cost constraint. For a smaller single site (e.g. a hotel with only 100 rooms), deploying a complete master-backed dual-airframe may be costly.Regular backup + quick recovery manual, to automatically back up data on a daily basis and quickly restore the environment on a new backup server when the main system is malfunctioning, using detailed operating manuals, while temporarily opening free access during recovery. Such options may be low but may take one or two hours, may not be acceptable for high-end sites and may be reasonably balanced for economically viable locations.
Consistency protection of billing dataWhen a server deals with a collection of billing records, which are written and which are not? Without a service protection mechanism, there may be duplication or leakage charges after restart. The system needs to use design such as scaffolding when the billing data is written to ensure that even if the same record is written several times, the end result is correct; and using the transaction log to determine clearly what records have been completed and what needs to be supplemented at recovery.
Disaster tolerance exercises are the most susceptible to being omitted from disaster design but are also the most important part of it. There is often a wide gap between the disaster-treatment programmes on file and those that can be properly implemented under pressure. It is recommended that they be performed quarterly or semi-annually: simulated main server crashes, operators switch to pre-cases, recording actual recovery times. Problems discovered through exercises (incidental steps, unskilled operators, backup data cannot be restored) must be updated to an operational manual after the formal exercise.
Cloud backup and local backup combination policyThe cost of a backup cloud has become very low as the price of cloud storage declines, even with small and medium-sized network billing systems deployed, adding a copy of a cloud is usually within tens of dollars per month, which can be ignored when losses are compared to lost data. In the absence of any particular reason, it should be defaulted to turn on the backup, rather than wait until data is lost before regretting it.
The redundancy of operator lines is also part of the disaster-resilient system.Network billing systemIn locations where network stability is more demanding, consideration could be given to the configuration of dual operators (the main and back-up links are from different operators) and the automatic switch to the back-up link in case of major chain failure. This option requires hardware-level support (double WAN routers or SD-WAN devices), with higher costs than a single-link scheme, but in a pay-for-fee scenario, each payment interruption would have a direct loss of revenue, which deserves careful assessment.
Design of disaster response and backup is not a technical problem, it is an engineering management issue. The technology options can be perfect, but without the accompanying operating protocols, responsible persons, and regular exercise mechanisms, they are likely to fail in real trouble.Network billing system(b) The critical dimensions of long-term operational capability.