Portal certification system is on the line, not the end of the project, but the start of a transport. The system runs smoothly and no one boasts it, so daily targeting of indicators is much more important than the moment we get online. When we make a hand-over to schools, we'll have to put a few core indicators in place, and teachers at the Internet center look at them every day.
Indicator I: Certification success rate
This is the most visible level of health. Normality should be at a flat height, and if it falls below a threshold, it may be a server, database or access device. We set success rates as one level of warning, so we can just go to people because it directly affects everyone ' s access to the Internet. The indicator board must be placed in the most visible position, not buried in the logs.
Indicator II: average certification time delay
Users say that the indexing loop is a rise. We usually give time-trips a baseline, like normal one or two seconds, beyond warning. Time lags are barometers of user experience, more sensitive than success rates, and often move before success rates drop.
Indicator III: Number of simultaneous online releases
Look how many devices are online, compared to the system's maximum capacity, and we suggest that Internet center teachers look at this amount before late peaks, if they're running out, pre-empting current or buildup, rather than waiting for it to collapse. The excess is a precautionary indicator, and its value is before it gets out of question.
Indicator IV: Servers and connections
The number of CPUs, memory and active connections to the authentication server is an hour before peak. We usually look at these links with the excess capacity, which are fast reaching the top, and the CPU running at full speed, as a clear maturation signal. This level of indicators is technologically biased, but it is the real reason behind those previous users' experiences.
Indicator V: Log writing and retention
The audit requires that the logs be kept continuously, and that an abnormal amount of log writing or a wrong life cycle be dealt with immediately. We have seen schools stop writing because they have disks full, and lose their records when they are checked. This line is better than keeping them open.
Indicator VI: abnormal login and shared hits
The unusual login and account-sharing hits detected by the system are security clues. We make them regular reports, we sweep every week in the network center, we find suspicious accounts before processing them, and it is not necessary to stop them in real time. This indicator is quiet, and a critical moment will help schools to block safety risks.
We need to grade the alarm.
Not all anomalies are called at midnight. We suggest that the alarm class: one that affects the whole network is sent to people directly, and the second category of small issues is dealt with only a few days later. Three weekly reports are simply statistically sufficient. The classing is bad, the alarm is heard, the teacher is numb, and no one takes it seriously.
Plan's going to be drilled.
One day, a real problem comes with the plan. The central database is downgraded to local level, payments are interrupted and refilled, certification services are full of how to limit the core of the current insurance, which is written in documents and one session at a time of practice.
Shut up: Watching the day to keep it small and big, pre-emptive.
The watchboards are for the right person.
The indicators are complete, and nobody looks at them. We suggest that we look at the board: first-line teachers see performance rates and time frames, the capacity and server of the supervisor, security guards watch abnormal hits. Different players view different views, alerts follow role lines, so that no one is inundated with irrelevant information, and people respond when real problems arise.
Do a little retrench once a week.
In addition to real-time attention, we suggest that the network center make a small retweets of surveillance data every week to see if the peak is innovative, which indicators are volatile, or whether there are warnings. The retraces do not take half an hour, but they can be detected in advance, so that one day they will suddenly fail. This habit is more valuable than buying more expensive equipment and it is a compound for transporting wiring.
Handover and file cannot be broken
The transporter is most afraid of the loss of experience. We stress that indicators, alerting and pre-emptive locations must be documented and handed over, so that new teachers can take over. Many schools are stuck in old carriers after they leave, and little problems are caused by a major breakdown.
Indicators are best tied to big screens or on-duty computers, so the whole picture is visible when you look up. We've seen schools that hide panels in their backstage menus, and nobody finds problems first. Visibility is a prevention in itself, exposing abnormally early, dealing with them earlier than remedying after.
Portal's daily operations are focused on success rates, delays, co-matches, servers, logs, anomalies, warning classification, and regular exercises.