diff --git a/README.gtt.txt b/README.gtt.txt
new file mode 100644
index 00000000000..d181df9acd7
--- /dev/null
+++ b/README.gtt.txt
@@ -0,0 +1,172 @@
+Global Temporary Table(GTT)
+=========================================
+
+Feature description
+-----------------------------------------
+
+Previously, temporary tables are defined once and automatically
+exist (starting with empty contents) in every session before using them.
+
+The temporary table implementation in PostgreSQL, known as Local temp tables(LTT),
+did not fully comply with the SQL standard. This version added the support of
+Global Temporary Table .
+
+The metadata of Global Temporary Table is persistent and shared among sessions.
+The data stored in the Global temporary table is independent of sessions. This
+means, when a session creates a Global Temporary Table and writes some data.
+Other sessions cannot see those data, but they have an empty Global Temporary
+Table with same schema.
+
+Like local temporary table, Global Temporary Table supports ON COMMIT PRESERVE ROWS
+or ON COMMIT DELETE ROWS clause, so that data in the temporary table can be
+cleaned up or preserved automatically when a session exits or a transaction COMMITs.
+
+Unlike Local Temporary Table, Global Temporary Table does not support
+ON COMMIT DROP clauses.
+
+In following paragraphs, we use GTT for Global Temporary Table and LTT for
+local temporary table.
+
+Main design ideas
+-----------------------------------------
+In general, GTT and LTT use the same storage and buffer design and
+implementation. The storage files for both types of temporary tables are named
+as t_backendid_relfilenode, and the local buffer is used to cache the data.
+
+The schema of GTTs is shared among sessions while their data are not. We build
+a new mechanisms to manage those non-shared data and their statistics.
+Here is the summary of changes:
+
+1) CATALOG
+GTTs store session-specific data. The storage information of GTTs'data, their
+transaction information, and their statistics are not stored in the catalog.
+
+2) STORAGE INFO & STATISTICS INFO & TRANSACTION INFO
+In order to maintain durability and availability of GTTs'session-specific data,
+their storage information, statistics, and transaction information is managed
+in a local hash table tt_storage_local_hash.
+
+3) DDL
+Currently, GTT supports almost all table'DDL except CLUSTER/VACUUM FULL.
+Part of the DDL behavior is limited by shared definitions and multiple copies of
+local data, and we added some structures to handle this.
+
+A shared hash table active_gtt_shared_hash is added to track the state of the
+GTT in a different session. This information is recorded in the hash table
+during the DDL execution of the GTT.
+
+The data stored in a GTT can only be modified or accessed by owning session.
+The statements that only modify data in a GTT do not need a high level of
+table locking. The operations making those changes include truncate GTT,
+reindex GTT, and lock GTT.
+
+4) MVCC commit log(clog) cleanup
+Each GTT in a session has its own piece of data, and they have their own
+transaction information. We set up data structures to track and maintain
+this information. The cleaning of CLOGs also needs to consider the transaction
+information of GTT.
+
+Detailed design
+-----------------------------------------
+
+1. CATALOG
+1.1 relpersistence
+define RELPERSISTENCEGLOBALTEMP 'g'
+Mark Global Temporary Table in pg_class relpersistence to 'g'. The relpersistence
+of indexes created on the GTT, sequences on GTT and toast tables on GTT are
+also set to 'g'
+
+1.2 on commit clause
+LTT's status associated with on commit DELETE ROWS and on commit PRESERVE ROWS
+is not stored in catalog. Instead, GTTs need a bool value on_commit_delete_rows
+in reloptions which is shared among sessions.
+
+1.3 gram.y
+GTT is already supported in syntax tree. We remove the warning message
+"GLOBAL is deprecated in temporary table creation" and mark
+relpersistence = RELPERSISTENCEGLOBALTEMP.
+
+2. STORAGE INFO & STATISTICS INFO & TRANSACTION INFO
+2.1. gtt_storage_local_hash
+Each backend creates a local hashtable gtt_storage_local_hash to track a GTT's
+storage file information, statistics, and transaction information.
+
+2.2 GTT storage file info track
+1) When one session inserts data into a GTT for the first time, record the
+storage info to gtt_storage_local_hash.
+2) Use beforeshmemexit to ensure that all files of session GTT are deleted when
+the session exits.
+
+2.3 statistics info
+1) relpages reltuples relallvisible relfilenode
+2) The statistics of each column from pg_statistic
+All the above information is stored in gtt_storage_local_hash.
+When doing vacuum or analyze, GTT's statistic is updated, which is used by
+the SQL planner.
+The statistics summarizes only data in the current session.
+
+2.3 transaction info track
+frozenxid minmulti from pg_class is stored to gtt_storage_local_hash.
+
+3 DDL
+3.1. active_gtt_shared_hash
+This is the hash table created in shared memory to trace the GTT files initialized
+in each session. Each hash entry contains a bitmap that records the backendid of
+the initialized GTT file. With this hash table, we know which backend/session
+is using this GTT. Such information is used during GTT's DDL operations.
+
+3.2 DROP GTT
+One GTT is allowed to be deleted when there is only one session using the table
+and the session is the current session. After holding the lock on GTT,
+active_gtt_shared_hash is checked to ensure that.
+
+3.3 ALTER GTT/DROP INDEX ON GTT
+Same as drop GTT.
+
+3.4 CREATE INDEX ON GTT
+1) create index on GTT statements build index based on local data in a session.
+2) After the index is created, record the index metadata to the catalog.
+3) Other sessions can enable or disable the local GTT index.
+
+3.5 TRUNCATE/REINDEX GTT
+The SQL truncate/reindex command open the GTT using AccessShareLock lock,
+not AccessExclusiveLock, because this command only cleans up local data and
+local buffers in current session. This allows these operations to be executed
+concurrently between sessions, unlike normal tables.
+
+3.6 LOCK GTT
+A lock GTT statement does not hold any relation lock.
+
+3.7 CLUSTER GTT/VACUUM FULL GTT
+The current version does not support.
+
+4 MVCC commit log(clog) cleanup
+
+The GTT storage file contains transaction information. Queries for GTT data rely
+on transaction information such as clog. The transaction information required by
+each session may be completely different. We need to ensure that the transaction
+information of the GTT data is not cleaned up during its lifetime and that
+transaction resources are recycled at the instance level.
+
+4.1 The session level GTT oldest frozenxid
+1) To manage all GTT transaction information, add session level oldest frozenxid
+in each session. When one GTT is created or removed, record the session level
+oldest frozenxid and store it in MyProc.
+2) When vacuum advances the database's frozenxid, session level oldest frozenxid
+should be considered. This is acquired by searching all of MyProc. This way,
+we can avoid the clog required by GTTs to be cleaned.
+
+4.2 vacuum GTT
+Users can perform vacuum over a GTT to clean up local data in the GTT.
+
+4.3 autovacuum GTT
+Autovacuum skips all GTTs, because the data in GTTs is only visible in current session.
+
+5 OTHERS
+5.1 Parallel query
+Planner does not produce parallel query plans for SQL related to GTT. Because
+GTT private data cannot be accessed across processes.
+
+5.2 WAL and Logical replication
+Like LTT, the DML on GTT does not record WAL and is not parsed or replay by
+the logical replication.
\ No newline at end of file
-- 
2.30.1 (Apple Git-130)