specificity of time series data for smart metering. For first

requirement when we have 2 million meters with 10 different parameters

(like active, passive power). New data come every 15 minutes which we

have to be stored on discs. Second requirement we have to store and

query for 1 year per meter. There is a lot of data to write and in the

same time we can expect hundreds queries per second. Each query can

request measurements for specific meter for day, month or year. With

such a requirements how data are organized on disc matters. You can ask

disc once or 365 * 96 times for one year query. I have tested many

relational DBs possibilities like various indexes and clustering types

and did not found solution. Finally, we have developed our own data base

(PulsarTSDB). Our functionality allow to configure how exactly data should

be stored and searched. Surprisingly, we found that

we able to meet this requirements on single commodity server.

To summarize, we have very specific non relational data, that needs

special treatment. The relational DBMS does not have functionality to

handle this data. Additionally, time series data does not come alone, but

are connected to the meters, customers, and contract handled in relational

DB. The historian data bases are integrated to sql data bases to

make life easer for users. We have an integration with Oracle DB in

contrast to OSIsoft.