- Ephemeris is a no-SQL in-memory software-only solution
- No specialized hardware is required
- Ephemeris runs on standard commodity servers and computers
- Cost savings are enormous vs. traditional database and server solutions
Most traditional databases round the space they reserve per field up to the next whole byte (or, often, the next power of two in bytes). Ephemeris goes down to the bit level. If we can store a field value in five bits we use just those five bits and we use the remaining three bits for something else. We pack the data very tightly and as a result Ephemeris can store much more data in the same amount of space without doing any compression.
Ephemeris doesn’t store individual cell values per record for some of the fields, instead it stores cell value combinations. In an address database for the US, for example, each ZIP-code will only occur in a small number of cities, one or two counties and in just one state. Instead of storing the ZIP, city, county and state as individual data for each record Ephemeris only stores the occurring combination of values for a record.
Ephemeris Does Not Use Compression
The file format stored on disk is (almost) exactly the same as the file format that we use when we run the data in memory. As a result data loading times are extremely fast because we do not have to do any conversions during file loads. Loading a database is (almost) a straight copy of the data from disk to RAM. The small footprint has resulted in another interesting advantage: access speeds. Because the Ephemeris’ file format is so tiny less memory needs to be accessed to find the records that you need to retrieve. Data access speeds in Ephemeris are in the 10 nanoseconds per record range on standard commodity servers. This means that you can traverse 100 million records per second per CPU core.
Read more about Ephemeris.