Aerospike-Architecture series of big data types
Large Data Types (Big Data Type) Big Data bin can store a Large number of object sets. The object type can be Integer/String/Map/List. A single record of Aerospike can contain many such bins. Such a bin allows a large amount of data to be contained in a separate record. The limit is based on the storage capability and there is no maximum value for the record.
The storage type used by Aerospike LDT, which is called sub-records ). Sub-records are similar to the conventional Aerospike records, except that they are connected to the parent record. The record shares the partition ID and internal record lock with the parent record. Therefore, the record is moved along with the parent record during migration and is protected along with the parent record through the same isolation mechanism.
Objects in LDT are not stored consecutively with their associated records, but are divided into sub-records (sizes: 2 kb to kb, determined by the size of the written block ). Sub-records are connected and indexed together and managed in the database through user-defined functions (UDFs. Using a sub-record means that accessing an object in LDT affects only a few sub-records rather than the entire record.
Large Ordered List (Large Ordered List)Large ordered list type (llist)Manage the list of large-scale sorting tasks. Objects in an ordered list can be of the original type (for example, integer, string, blob) or complex data types (for example, map, list)
Figure 5a large ordered list operation
Physically, a large sequence table is implemented by the B + tree. The node and leaf of the tree are sub-records ). The root of the B + tree resides in the parent record.
Figure 5b large ordered list Physical Layer
There is no depth limit on the B + tree structure, so there is no limit on the number of objects stored in the large ordered list.
The LLIST control structure contains all metadata information about a large ordered list. For example, it includes the following information: