Access Scene of Data Analyst

In the data analysis scenario, data analyst often uses 「10-minute mean」or「15-minute mean」as sample data for analysis, but does not use the original acquisition data directly. If there is only original acquisition data in the cloud platform, data analyst usually gets the「10-minute mean」by use of the following methods:

1. Export the original data, and obtain the 10-minute mean via data preprocessing tool.
2. Add stream computing task in the cloud platform, calculate the 10-minute mean and save it in the database timely.
3. Perform the offline tasks to historical data via big data platform, calculate the 10-minute mean of historical data, and save it in the database.

Through the solutions above, the data analyst can get the 10-minute mean, however, these solutions have a very high cost and they are inconvenient.

The problem above can be solved better through edge stream computing. The edge node gets 10-minute mean via stream computing on the edge side, then the mean is reported to the cloud iothub, and 10min data is transferred to database via rule engine, to greatly reduce the access difficulty of data analyst.

Operation and Maintenance Staff’s Real-time Monitoring Scenario

In the Internet of Things scenario, device measuring data often jitters due to various factors (network factor, and device accuracy factor). If the device real-time acquisition value is configured with threshold alarm, false alarm will often occur, resulting in that lots of useless alarms need to be handled, and you will lose confidence in the accuracy of alarm gradually, so threshold alarm performs practically no function.

For this scenario, you may reduce the deviation caused by data jitter by virtue of stream computing ability, and the common solutions include:

1. Alarm by mean: Get 10-minute mean, 10-minute maximum, 10-minute minimum, and 10-minute computation sample number by stream computing, and then set threshold rules, for example, “10-minute mean>threshold and 10-minute computation sample number>100”.

2. Alarm by duration: Find out the device having the real-time acquisition value always greater than the specified threshold and lasting for a very long time, for example, "device temperature>100℃ and duration>5minutes".

   In the case of network instability, if the two solutions above are realized via cloud stream computing, the finally obtained computation values have a low accuracy, for example, when the device is out of connection for 5 minutes, 10-minute mean is computed in the cloud, therefore, the computation result is not accurate. The edge side is intranet environment, so the probability of network anomaly is greatly reduced. Calculate the statistical value on the edge side and report it to the cloud, to greatly improve the accuracy of stream computing statistical result.

Edge Stream Computing Value

• Reduce the cost, including traffic cost, storage cost, and cloud stream computing resource cost.
• In weak network environment, improve the accuracy of stream computing result.

Related Links

Please see Operation Guide for use of edge stream computing.