Storage Formats
As part of Phoenix 4.10, we have reduced the on disk storage size to improve overall performance by implementing the following enhancements:
- Introduce a layer of indirection between Phoenix column names and the corresponding HBase column qualifiers.
- Support a new encoding scheme for immutable tables that packs all values into a single cell per column family
How to use column mapping?
One can set the column mapping property only at the time of creating the table. Before deciding on using column mapping, you need to think about how many columns you expect in a table and its view hierarchy to have in the lifecycle. For various mapping schemes, below are the limits on number of columns:
Config/Property Value | Max # of columns |
1 | 255 |
2 | 65535 |
3 | 16777215 |
4 | 2147483647 |
NONE | no limit(theoretically) |
For mutable tables, this limit applies to columns in all column families. For immutable tables, the limit applies to per column family.By default, any new phoenix tables will be using the column mapping feature.
These defaults could be overridden by setting below config to the desired value in hbase-site.xml
Table type | Default Column mapping | Config |
Mutable/Immutable | 2 byte qualifiers | phoenix.default.column.encoded.bytes.attrib |
Keep in mind that this config controls global level defaults that would apply to all tables. If you would like to use a different mapping scheme than this global default, then you can use the COLUMN_ENCODED_BYTES table property.
CREATE TABLE T
(
a_string varchar not null,
col1 integer
CONSTRAINT pk PRIMARY KEY (a_string)
)
COLUMN_ENCODED_BYTES = 1;
How to use immutable data encoding?
Below are some scenarios in when it would be better to use ONE_CELL_PER_COLUMN encoding instead.
- Data is sparse i.e. less than 50% of the columns have values
- Size of data within a column family gets too big. Our general guidance here is that with default HBase block size of 64K, if data within a column family grows beyond 50K then we shouldn’t be using SINGLE_CELL_ARRAY_WITH_OFFSETS.
- For immutable tables that are going to have views on them
SINGLE_CELL_ARRAY_WITH_OFFSETS generally provides really good performance improvement and space savings
By default, immutable non-multitenant tables are created using the two byte column mapping and the SINGLE_CELL_ARRAY_WITH_OFFSETS data encoding.
Immutable Table type | Immutable storage scheme | Config |
Multi-tenant | ONE_CELL_PER_COLUMN | phoenix.default.multitenant.immutable.storage.scheme |
Non multi-tenant | SINGLE_CELL_ARRAY_WITH_OFFSETS | phoenix.default.immutable.storage.scheme |