\chapter*{Abstract}

Relational DataBase(RDB) has problem of impedance mismatch.
Exist an OR Mapper that can use database records as objects in program.
and DB has been expand such as table specialization KVS and Json correspondence.
However program Construction complexity structure on memory.
It has gap between  data in program and data in database.

Laboratory develops database Jungle that resolves these problems.
Jungle doesn't destroy the tree structure.
Construction a new tree while saving trees.
Jungle transaction by atomic exchange the root of the tree.
Program can use Jungle tree and not ask database.
Jungle jas Consistency for every tree.
Possible create new consistent tree by merge consistent trees together.
Easy to send data.

Jungle can read fast.
However writing speed depend on Structure of tree and tree size.
worst case Order(n).
And create Index is slow.
This research will improve Jungle editing function.
After, Develop and use applications using Jungle.

benchmark was confirm the speedup of the tree editing function.
And Jungle able to read data faster than PostgreSQL and MongDB.