Inside PostgreSQL Shared Memory
BRUCE MOMJIAN,
ENTERPRISEDB
October, 2008
Abstract
POSTGRESQL is an open-source, full-featured relational database. This
presentation gives an overview of the shared memory structures used
by Postgres.
Outline
1. File storage format
2. Shared memory creation
3. Shared buffers
4. Row value access
5. Locking
6. Other structures
Inside PostgreSQL Shared Memory 1
File System /data
Postgres /data
Postgres
Postgres
Inside PostgreSQL Shared Memory 2
File System /data/base
Postgres /data /base
/global
Postgres /pg_clog
/pg_multixact
/pg_subtrans
Postgres /pg_tblspc
/pg_twophase
/pg_xlog
Inside PostgreSQL Shared Memory 3
File System /data/base/db
Postgres /data /base /16385 (production)
/1 (template1)
Postgres /16821 (test)
/17982 (devel)
/21452 (marketing)
Postgres
Inside PostgreSQL Shared Memory 4
File System /data/base/db/table
Postgres /data /base /16385 /24692 (customer)
/27214 (order)
Postgres /25932 (product)
/25952 (employee)
/27839 (part)
Postgres
Inside PostgreSQL Shared Memory 5
File System Data Pages
Postgres /data /base /16385 /24692
8k 8k 8k 8k
Postgres
Postgres
Inside PostgreSQL Shared Memory 6
Data Pages
Postgres /data /base /16385 /24692
8k 8k 8k 8k
Postgres
Postgres
Page Header Item Item Item
8K
Tuple
Tuple Tuple Special
Inside PostgreSQL Shared Memory 7
File System Block Tuple
Postgres /data /base /16385 /24692
8k 8k 8k 8k
Postgres
Page Header Item Item Item
Postgres
8K
Tuple
Tuple Tuple Special
Tuple
Inside PostgreSQL Shared Memory 8
File System Tuple
int4in(’9241’) ’Martin’
Tuple
textout()
Header Value Value Value Value Value Value
OID − object id of tuple (optional)
xmin − creation transaction id
xmax − destruction transaction id
cmin − creation command id
cmax − destruction command id
ctid − tuple id (page / item)
natts − number of attributes
infomask − tuple flags
hoff − length of tuple header
bits − bit map representing NULLs
Inside PostgreSQL Shared Memory 9
Tuple Header C Structures
typedef struct HeapTupleFields
{
TransactionId t_xmin; /* inserting xact ID */
TransactionId t_xmax; /* deleting or locking xact ID */
union
{
CommandId t_cid; /* inserting or deleting command ID, or both */
TransactionId t_xvac; /* VACUUM FULL xact ID */
} t_field3;
} HeapTupleFields;
typedef struct HeapTupleHeaderData
{
union
{
HeapTupleFields t_heap;
DatumTupleFields t_datum;
} t_choice;
ItemPointerData t_ctid; /* current TID of this or newer tuple */
/* Fields below here must match MinimalTupleData! */
uint16 t_infomask2; /* number of attributes + various flags */
uint16 t_infomask; /* various flag bits, see below */
uint8 t_hoff; /* sizeof header incl. bitmap, padding */
/* ^ − 23 bytes − ^ */
bits8 t_bits[1]; /* bitmap of NULLs −− VARIABLE LENGTH */
/* MORE DATA FOLLOWS AT END OF STRUCT */
} HeapTupleHeaderData;
Inside PostgreSQL Shared Memory 10
Shared Memory Creation
k()
for
postmaster postgres postgres
Program (Text) Program (Text) Program (Text)
Data Data Data
Shared Memory Shared Memory Shared Memory
Stack Stack Stack
Inside PostgreSQL Shared Memory 11
Shared Memory
PROC Lightweight Locks XLOG Buffers
Proc Array Lock Hashes CLOG Buffers
LOCK Subtrans Buffers
Auto Vacuum PROCLOCK Two−Phase Structs
Btree Vacuum Multi−XACT Buffers
Free Space Map Statistics
Background Writer Synchronized Scan Shared Invalidation
Buffer Descriptors
Shared Buffers
Semaphores
Inside PostgreSQL Shared Memory 12
Shared Buffers
Buffer Descriptors Pin Count − prevent page replacement
LWLock − for page changes
8k 8k 8k
Shared Buffers
read()
Page Header Item Item Item
write()
Postgres /data /base /16385 /24692
8K
8k 8k 8k 8k
Tuple
Postgres
Tuple Tuple Special
Postgres
Inside PostgreSQL Shared Memory 13
HeapTuples
8k 8k 8k
Shared Buffers
Page Header Item Item Item
8K
Tuple
Tuple Tuple Special
HeapTuple
int4in(’9241’) ’Martin’
Tuple
textout()
Header Value Value Value Value Value Value Postgres
C pointer
OID − object id of tuple (optional)
xmin − creation transaction id
xmax − destruction transaction id
cmin − creation command id
cmax − destruction command id
ctid − tuple id (page / item)
natts − number of attributes
infomask − tuple flags
hoff − length of tuple header
bits − bit map representing NULLs
Inside PostgreSQL Shared Memory 14
Finding A Tuple Value in C
Datum
nocachegetattr(HeapTuple tuple,
int attnum,
TupleDesc tupleDesc,
bool *isnull)
{
HeapTupleHeader tup = tuple−>t_data;
Form_pg_attribute *att = tupleDesc−>attrs;
{
int i;
/*
* Note − This loop is a little tricky. For each non−null attribute,
* we have to first account for alignment padding before the attr,
* then advance over the attr based on its length. Nulls have no
* storage and no alignment padding either. We can use/set
* attcacheoff until we reach either a null or a var−width attribute.
*/
off = 0;
for (i = 0;; i++) /* loop exit is at "break" */
{
if (HeapTupleHasNulls(tuple) && att_isnull(i, bp))
continue; /* this cannot be the target att */
if (att[i]−>attlen == −1)
off = att_align_pointer(off, att[i]−>attalign, −1,
tp + off);
else
/* not varlena, so safe to use att_align_nominal */
off = att_align_nominal(off, att[i]−>attalign);
if (i == attnum)
break;
off = att_addlength_pointer(off, att[i]−>attlen, tp + off);
}
}
return fetchatt(att[attnum], tp + off);
}
Inside PostgreSQL Shared Memory 15
Value Access in C
#define fetch_att(T,attbyval,attlen) \
( \
(attbyval) ? \
( \
(attlen) == (int) sizeof(int32) ? \
Int32GetDatum(*((int32 *)(T))) \
: \
( \
(attlen) == (int) sizeof(int16) ? \
Int16GetDatum(*((int16 *)(T))) \
: \
( \
AssertMacro((attlen) == 1), \
CharGetDatum(*((char *)(T))) \
) \
) \
) \
: \
PointerGetDatum((char *) (T)) \
)
Inside PostgreSQL Shared Memory 16
Test And Set Lock
Can Succeed Or Fail
1 1
0/1
0 1
Success Failure
Was 0 on exchange Was 1 on exchange
Lock already taken
Inside PostgreSQL Shared Memory 17
Test And Set Lock
x86 Assembler
static __inline__ int
tas(volatile slock_t *lock)
{
register slock_t _res = 1;
/*
* Use a non−locking test before asserting the bus lock. Note that the
* extra test appears to be a small loss on some x86 platforms and a small
* win on others; it’s by no means clear that we should keep it.
*/
__asm__ __volatile__(
" cmpb $0,%1 \n"
" jne 1f \n"
" lock \n"
" xchgb %0,%1 \n"
"1: \n"
: "+q"(_res), "+m"(*lock)
:
: "memory", "cc");
return (int) _res;
}
Inside PostgreSQL Shared Memory 18
Spin Lock
Always Succeeds
1 1
0/1 Sleep of increasing duration
0 1
Success Failure
Was 0 on exchange Was 1 on exchange
Lock already taken
Spinlocks are designed for short-lived locking operations, like
access to control structures. They are not be used to protect
code that makes kernel calls or other heavy operations.
Inside PostgreSQL Shared Memory 19
Light Weight Locks
Sleep On Lock
PROC Lightweight Locks XLOG Buffers
Proc Array Lock Hashes CLOG Buffers
LOCK Subtrans Buffers
Auto Vacuum PROCLOCK Two−Phase Structs
Btree Vacuum Multi−XACT Buffers
Free Space Map Statistics
Background Writer Synchronized Scan Shared Invalidation
Buffer Descriptors
Shared Buffers
Semaphores
Light weight locks attempt to acquire the lock, and go to
sleep on a semaphore if the lock request fails. Spinlocks
control access to the light weight lock control structure.
Inside PostgreSQL Shared Memory 20
Database Object Locks
PROC PROCLOCK LOCK
Lock Hashes
Inside PostgreSQL Shared Memory 21
Proc
PROC
empty used used empty used empty
Proc Array
Inside PostgreSQL Shared Memory 22
Other Shared Memory Structures
PROC Lightweight Locks XLOG Buffers
Proc Array Lock Hashes CLOG Buffers
LOCK Subtrans Buffers
Auto Vacuum PROCLOCK Two−Phase Structs
Btree Vacuum Multi−XACT Buffers
Free Space Map Statistics
Background Writer Synchronized Scan Shared Invalidation
Buffer Descriptors
Shared Buffers
Semaphores
Inside PostgreSQL Shared Memory 23
Conclusion
Pink Floyd: Wish You Were Here
Inside PostgreSQL Shared Memory 24
