I want to create job which watch the number of record exception in series of specific time of days. It means the job will compare the number exception in 6pm of each day .My team usually upcode in 2 or 3am, so by comparing the number of exception in that day with these old days, we can detect the problem.
Well I would say the best thing to do is just run [high_/low_]count detection. We automatically model seasonality in the signal and so if you have different rates of exceptions for different times of day we will be comparing the count at that time of day to the count at the same time on previous days. It does take us a little while to detect daily seasonality (but usually 3 days is sufficient).
What I meant was we should learn that there is daily seasonality after observing about 3 days of data, it can be a bit longer depending on the data characteristics but this is usually sufficient. This is independent of bucket length although if you use a very long bucket length, such as 1 day or more we wouldn't model daily seasonality.
Apache, Apache Lucene, Apache Hadoop, Hadoop, HDFS and the yellow elephant
logo are trademarks of the
Apache Software Foundation
in the United States and/or other countries.