v0.5.0

• True columnar database - each time-series stored in dedicated file.
• Automatic data erasing was disabled. Timeseries removing only by manual.
• UTC time in inner functions.
• Golang http client
• Interval aggregation support. Available intervals: raw,minute, half hour, hour, day, week, month, year.

v0.4.1: Statistical functions

• google.test for unittests.
• statistical functions:
  • average
  • median
  • sigma(standard deviation)
  • percentile90
  • percentile99

v0.4.0

• Data compaction methods.
• Less allocations in read queries.
• Auto erase old values.
• Magic numbers in page file.
• Settings:
  • max_chunks_per_page - maximum count chunks per page when write from memory storage.
  • max_store_period - period, when old pages must be removed.
• Wal now contained in-memory index.
• Rename build options: add prefix 'DARIADB_'
• Memory only mode.
• New memory storage.
• Full featured http api.
• Network client now can work with scheme.

v0.3.0

• Remove bystep storage engine.
• Accept unordered data.
• Statistic:
  • time min/max
  • value min/max
  • measurement count
  • values sum
• Page can be repacked.
• Scheme API.
• Implement shard-storage.

v0.2.1

code refactoring.

v0.2.0

• Full featured server with client library(C++).
• Engine version in manifest.
• Engine write strategies.
• New algorithm to drop values between levels depending on the strategies.
• New logger interface.
• Now can remove old compressed pages.
• New compression algorithm: speed increased by 3 times.
• Flag 'source' was removed from measurement.
• Measurement Id is uint32_t.
• Remove COLA layer.
• Speed up dropping data between layers.
• Old pages compaction.
• Full futured memory storage.
• cache-mode - all values stored in memory with async writes to disk.
• Lockfile in storage.
• New storage control tool.
• SQLite3 based manifest.
• Foreach - is async.
• Bystep storage.
• Remove mmap usage.
• Different folders for 'raw' and 'bystep' values.

v0.1.1

• Code refactoring.
• Compressed level is use "append only" idiom. More crash safety and more faster.
• Use Jenkins hash function for bloom filter.
• Apache License.
• Options file in storage.

Asynchronous storage.

• chromium thread model (actors).
• async storage.
• more crash safety.
• map (lie map/redice) api.

LSM-like struct.

• Accept unordered data.
• Restore after crash.
• CRC32 checksum for data.
• LSM-like struct: COLA layer, Append-only files layer, Compressed layer.
• Bloom filter based on std::hash.

file storage.

• base-COLA for sort new measurements.
• new compressed storage.
• LEB128 compression for Id and Flag.
• LRU memory cache.