Many sites have more than one web export, some with broadband from many operators, and some with two main lines. Multiple exports raise bandwidth and reliability, but it creates an additional problem for Portal certification systems: which line the relevant flow is and what happens if it goes wrong. This problem does not exist in a single-export environment; if it is not specifically addressed in a multi-export environment, there are peculiarities of users being certified when they do good or how they fail to find regular patterns.
The authentication traffic is separate from the Internet traffic.
The first is to distinguish between two types of traffic. One is the business flow generated by users online, which can be strategically distributed across different exports and anywhere. The other is to verify related flows, including requests for user-end access to authentication pages, interactions between access devices and authentication platforms, and release and strategy instructions from the platform. The latter type requires much higher levels of stability on the path, and certification fails or slows down once it has gone an unstable route or changes back and forth.
The path changes back and forth are the greatest danger.
The most common mode of failure in multiple export environments is that the path is inconsistent. Users request to go out from one line, or to respond to another, or to walk back from another for a while or for a period of time. For ordinary Internet traffic, this asymmetry usually does not affect experience, but it is not right for such regular interactions, which can cause the platform to see a change in its origin address and earlier meetings.
Fixed exports to certified flows
The safest way to handle this is to bind the flow associated with certification to a fixed export. This is by designating, in an export strategy, a fixed line for the location of the authentication platform and the address where the access device interacts with the platform. This means that the path of authentication remains stable regardless of how the business flow is shared. After binding, it is verified that the certificate is normal when the line is normally certified and not affected when other lines are switched or shaking.
Consider session when main switch is made
Multiple exports are usually accompanied by master or load-sharing mechanisms, which match the route quality tests. Here, attention is paid to the effect of switching on established conversations: if the online user’s conversation will be interrupted after a line switch, need not be recertified, and whether it will shock the system because of sudden increases. These are all identified and validated at the configuration stage. If the switch will lead to large numbers of users being re-certified simultaneously, then the instant stress system will have to be evaluated as unmanageable, considering staggered or proactive in advance.
Switch threshold to actual mass
The route quality test is the basis of choice, but these parameters are determined by what, how often they are tested and what conditions are changed. Too frequent testing places a burden on the chain itself, too thin detection does not detect short-term deterioration. Switching thresholds are too sensitive to allow for frequent round-trips, while each switch may affect online users; too slow setting, which is already poor, and user experience first. Parameters need to be adjusted within the actual fluctuation of the line and cannot copy default values.
Backup links to reserve bandwidth
The most important aspect of the multi-export environment is the capacity of the backup links. Many projects have a much smaller provisioning bandwidth than the main line, which is not normally distributed at all, and when the main line fails to reach the main line, it is fully packed, with worse results than a single export. So it was designed to make clear how many traffics the main line would carry, whether full or only critical. If only key operations are insured, pre-positioned down strategies should be used to limit non-critical flows to major lines rather than crowding them together.
You need to check for a switch.
The multi-export items must have a switch test in the checklist: artificially letting main lines out, observing whether systems are going to be switched as expected, whether certification is still normal, how much online users are affected, and if they can automatically be returned after recovery. This test seems simple, but it can expose once and for all the configuration problems of route selection strategies, session processing, bandwidth reservation. Without this testing, these issues will remain buried until a real circuit breaks together.
The certification platform must be accessible at all exits.
The multi-export environment has a prerequisite to be verified: the address of the authentication platform, which can be accessed from every export. If, for tactical reasons, the platform address is only accessible from a certain route, if it fails, the certification will fail as a whole and the multiple export reliability advantage will expire completely.
Multiple-export logs and retrospectives
Finally, one thing that is easily missed: multiple exports affect the integrity of the logs. The same user flows may go from different exports and address conversion records vary from one export to another. If follow-up action is to be done by way of a back-to-back Internet connection, users, time, address, port and export information must be recorded together and recalibrated after the event, otherwise half of the records found out two paragraphs are sorry and will be cut off retroactively. This item is presented at a very low cost during the design phase and will be difficult to refill when it comes online.