Java Cache Data before entering to database

Question

we have to implement a monitoring module to existing web application. The purpose of the application is to record time taken at each method executed at run time and saving to 2 monitoring tables. As there are many method invocations happening in the application, we can't save to table every time a method is executed.We have used spring aop to intercept time.We used redis to cache data, and in every ten minutes, take the data from cache and save in the database.But redis seems to be a troublesome ideas as it keep making new connections and this give the application a nightmare.Is there any alternative way to do this.We considered about writing to a file and taking data from the file periodically.But that also seemed to be a resource consuming solutions.

• user207421: Surely your database can log the time for every transaction?
• Paul Vargas: Are you looking for speed or less consumption of resources? Use the RAM is always faster than using the hard disk. In another hand, you can try another caching solutions.
• ghTvNath: @PaulVargas,I think Its good idea.
• ghTvNath: @EJP idea is one table records all the transaction data, like time taken at each method and other table keeps computational record like avg,max,min time for one specific method.first table data will be deleted periodically
• user207421: I understand the idea. The point is that all the data is already in the database. All you have to do is perform queries on the logs. You don't have to collect the data.

Answer 1

I would suggest storing everything in your Java application. If all you want is a few stats, you can keep track of this information in a HashMap and write it every 10 minutes to the database from another thread. Min/Max/Avg time won't take much memory, and you can reset stats once it's written to the database.

If that doesn't suite your needs for some reason, Redis should be fast, and can even help you compute your statistics. I'd setup a connection pool to prevent it from making a new network connection each time. Then, I'd use asynchronous writes (pipelining) in your Redis client to speed up writes. Jedis and JRedis both support this.