2.8 KiB
2.8 KiB
| title | menu | ||||||
|---|---|---|---|---|---|---|---|
| Design insights and tradeoffs in InfluxDB |
|
InfluxDB is a time-series database. Optimizing for this use-case entails some tradeoffs, primarily to increase performance at the cost of functionality. Below is a list of some of those design insights that lead to tradeoffs:
- For the time series use case, we assume that if the same data is sent multiple times, it is the exact same data that a client just sent several times.
- Pro: Simplified conflict resolution increases write performance
- Con: Cannot store duplicate data; may overwrite data in rare circumstances
- Deletes are a rare occurrence. When they do occur it is almost always against large ranges of old data that are cold for writes.
- Pro: Restricting access to deletes allows for increased query and write performance
- Con: Delete functionality is significantly restricted
- Updates to existing data are a rare occurrence and contentious updates never happen. Time series data is predominantly new data that is never updated.
- Pro: Restricting access to updates allows for increased query and write performance
- Con: Update functionality is significantly restricted
- The vast majority of writes are for data with very recent timestamps and the data is added in time ascending order.
- Pro: Adding data in time ascending order is significantly more performant
- Con: Writing points with random times or with time not in ascending order is significantly less performant
- Scale is critical. The database must be able to handle a high volume of reads and writes.
- Pro: The database can handle a high volume of reads and writes
- Con: The InfluxDB development team was forced to make tradeoffs to increase performance
- Being able to write and query the data is more important than having a strongly consistent view.
- Pro: Writing and querying the database can be done by multiple clients and at high loads
- Con: Query returns may not include the most recent points if database is under heavy load
- Many time series are ephemeral. There are often time series that appear only for a few hours and then go away, e.g. a new host that gets started and reports for a while and then gets shut down.
- Pro: InfluxDB is good at managing discontinuous data
- Con: Schema-less design means that some database functions are not supported e.g. there are no cross table joins
- No one point is too important.
- Pro: InfluxDB has very powerful tools to deal with aggregate data and large data sets
- Con: Points don't have IDs in the traditional sense, they are differentiated by timestamp and series