Time series data is partitioned on two levels of storage groups and time ranges.
The storage group is specified by the user display. Use the statement “SET STORAGE GROUP TO” to specify the storage group. Each storage group has a corresponding StorageGroupProcessor.
The main fields it has are:
Read-write lock: insertLock
Unclosed sequential file processors for each time partition: workSequenceTsFileProcessors
Unclosed out-of-order file processor corresponding to each time partition: workUnsequenceTsFileProcessors
Full sequential file list for this storage group (sorted by time): sequenceFileTreeSet
List of all out-of-order files for this storage group (unordered): unSequenceFileList
A map that records the last write time of each device. When sequential data is flashed, the time recorded by this map is used: latestTimeForEachDevice
A map that records the last flash time of each device to distinguish between sequential and out-of-order data: latestFlushedTimeForEachDevice
A version generator map corresponding to each time partition, which is convenient for determining the priority of different chunks when querying: timePartitionIdVersionControllerMap
The data in the same storage group is partitioned according to the time range specified by the user. The related parameter is partition_interval and the default is week. That is, data of different weeks will be placed in different partitions.
StorageGroupProcessor performs partition calculation on the inserted data to find the specified TsFileProcessor, and the TsFile corresponding to each TsFileProcessor will be placed in a different partition folder.
The file structure after partitioning is as follows:
data
-- sequence
---- [Storage group name1]
------ [Time division ID1]
-------- xxxx.tsfile
-------- xxxx.resource
------ [Time division ID2]
---- [Storage group name 2]
-- unsequence