TS.CREATE

is key name for the time series.

is maximum age for samples compared to the highest reported timestamp, in milliseconds. Samples are expired based solely on the difference between their timestamp and the timestamps passed to subsequent TS.ADD, TS.MADD, TS.INCRBY, and TS.DECRBY calls with this key.

When set to 0, samples never expire. When not specified, the option is set to the global RETENTION_POLICY configuration of the database, which by default is 0.

specifies the series samples encoding format as one of the following values:

• COMPRESSED, applies compression to the series samples.
• UNCOMPRESSED, keeps the raw samples in memory. Adding this flag keeps data in an uncompressed form.

COMPRESSED is almost always the right choice. Compression not only saves memory but usually improves performance due to a lower number of memory accesses. It can result in about 90% memory reduction. The exception are highly irregular timestamps or values, which occur rarely.

When not specified, the option is set to COMPRESSED.

is initial allocation size, in bytes, for the data part of each new chunk. Actual chunks may consume more memory. Changing chunkSize (using TS.ALTER) does not affect existing chunks.

Must be a multiple of 8 in the range [48 .. 1048576]. When not specified, it is set to the global CHUNK_SIZE_BYTES configuration of the database, which by default is 4096 (a single memory page).

Note: Before v1.6.10 no minimum was enforced. Between v1.6.10 and v1.6.17 and in v1.8.0 the minimum value was 128. Since v1.8.1 the minimum value is 48.

The data in each key is stored in chunks. Each chunk contains header and data for a given timeframe. An index contains all chunks. Iterations occur inside each chunk. Depending on your use case, consider these tradeoffs for having smaller or larger sizes of chunks:

• Insert performance: Smaller chunks result in slower inserts (more chunks need to be created).
• Query performance: Queries for a small subset when the chunks are very large are slower, as we need to iterate over the chunk to find the data.
• Larger chunks may take more memory when you have a very large number of keys and very few samples per key, or less memory when you have many samples per key.

If you are unsure about your use case, select the default.

is policy for handling insertion (TS.ADD and TS.MADD) of multiple samples with identical timestamps, with one of the following values:

• BLOCK: ignore any newly reported value and reply with an error
• FIRST: ignore any newly reported value
• LAST: override with the newly reported value
• MIN: only override if the value is lower than the existing value
• MAX: only override if the value is higher than the existing value
• SUM: If a previous sample exists, add the new sample to it so that the updated value is equal to (previous + new). If no previous sample exists, set the updated value equal to the new value.

When not specified: set to the global DUPLICATE_POLICY configuration of the database (which, by default, is BLOCK).

BLOCK is often used to avoid accidental changes. FIRST can be used as an optimization when duplicate reports are possible. LAST can be used when updates are being reported. SUM is used for counters (e.g., the number of cars entering a parking lot per minute when there are multiple reporting counting devices). MIN and MAX can be used, for example, to store the minimal/maximal stock price per minute (instead of storing all the samples and defining a compaction rule).

is the policy for handling duplicate samples. A new sample is considered a duplicate and is ignored if the following conditions are met:

• The time series is not a compaction;
• The time series' DUPLICATE_POLICY IS LAST;
• The sample is added in-order (timestamp ≥ max_timestamp);
• The difference of the current timestamp from the previous timestamp (timestamp - max_timestamp) is less than or equal to IGNORE_MAX_TIME_DIFF;
• The absolute value difference of the current value from the value at the previous maximum timestamp (abs(value - value_at_max_timestamp) is less than or equal to IGNORE_MAX_VAL_DIFF.

where max_timestamp is the timestamp of the sample with the largest timestamp in the time series, and value_at_max_timestamp is the value at max_timestamp.

When not specified: set to the global IGNORE_MAX_TIME_DIFF and IGNORE_MAX_VAL_DIFF, which are, by default, both set to 0.

is set of label-value pairs that represent metadata labels of the key and serve as a secondary index.

The TS.MGET, TS.MRANGE, and TS.MREVRANGE commands operate on multiple time series based on their labels. The TS.QUERYINDEX command returns all time series keys matching a given filter based on their labels.