Hi
all,
Why
would anybody care about all these heap stuff? How does it help
performance....
Thanks.
Abraham Guerra
-----Original Message-----
From: Tanel Poder [mailto:[EMAIL PROTECTED] Sent: Tuesday, January 06, 2004 10:04 AM To: Multiple recipients of list ORACLE-L Subject: Re: (long) Top level heaps/subheaps Well, a heap is just a bunch of memory (allocated
in contiguous extents) and you can practically allocate any size of chunk
of memory from it (with some restrictions). A regular heap has freelist and LRU
list mechanisms for managing space in it as well as memory
allocation/releasing routines which you can then call. Oracle uses its generic
heap manager code (KGH) for managing it's memory.
Once you allocate a chunk from heap it's up to the allocator what he does with it. It can just use it for some kind of static data, but can also create another heap in these chunks it allocated, this heap may use different heap manager code for managing its memory if needed (for example shared pool reserved area and large pool behave a little differently, they don't have LRU list mechanisms). If you take any heapdump using alter session
set events 'immediate trace name heapdump level x' ; then top level heaps show
parent=00000000 in their descriptors. Note that this "top call heap" has
allocated chunks for heap with descriptor address 06B51DB0
(callheap) and also you see that "callheap" has a parent heap with
descriptor address 06B52790.
HEAP DUMP heap name="top call
heap" desc=06B52790
extent sz=0x213c alt=92 het=32767 rec=0 flg=2 opc=2 parent=00000000 owner=00000000 nex=00000000 xsz=0xfffc EXTENT 0 addr=071D0004 Chunk 71d000c sz= 112 perm "perm " alo=112 Chunk 71d007c sz= 65412 free " " EXTENT 1 addr=07170004 Chunk 717000c sz= 232 perm "perm " alo=232 Chunk 71700f4 sz= 60724 free " " Chunk 717ee28 sz= 1352 freeable "callheap " ds=06B51DB0 Chunk 717f370 sz= 1072 freeable "callheap " ds=06B51DB0 Chunk 717f7a0 sz= 1072 recreate "callheap " latch=00000000 ds 6b51db0 sz= 3496 ct= 3 [snipped....]
HEAP DUMP heap name="callheap" desc=06B51DB0 extent sz=0x424 alt=32767 het=32767 rec=0 flg=2 opc=3 parent=06B52790 owner=00000000 nex=00000000 xsz=0x53c EXTENT 0 addr=0717EE34 Chunk 717ee3c sz= 1332 freeable "qkkele " EXTENT 1 addr=0717F37C Chunk 717f384 sz= 572 perm "perm " alo=388 Chunk 717f5c0 sz= 444 free " " Chunk 717f77c sz= 36 freeable "qkkkey " EXTENT 2 addr=0717F7BC Chunk 717f7c4 sz= 28 perm "perm " alo=28 Chunk 717f7e0 sz= 956 free " " Chunk 717fb9c sz= 52 freeable "event string " Total heap size = 3420 FREE LISTS: Bucket 0 size=76 Bucket 1 size=268 Chunk 717f7e0 sz= 956 free " " Chunk 717f5c0 sz= 444 free " " Bucket 2 size=1036 Total free space = 1400 UNPINNED RECREATABLE CHUNKS (lru first): PERMANENT CHUNKS: Chunk 717f384 sz= 572 perm "perm " alo=388 Chunk 717f7c4 sz= 28 perm "perm " alo=28 Permanent space = 600 ****************************************************** If making these "full" heapdumps puts too much strain on your system, then it might be easier to use event HEAPDUMP_ADDR which allows to dump heaps only with specified descriptor address. There are other options as well, such is X$KSMHP table to query any heap (which is, blah, highly unsupported) or attaching directly to SGA shared memory segment and reading from there. Sorry for too long post, I'm currently too lazy to do
anything useful...
Tanel.
----- Original Message ----- From: Sultan Syed To: Multiple recipients of list ORACLE-L Sent: Tuesday, January 06, 2004 3:49 PM Subject: Top level heaps/subheaps Hi, I am getting little confussion how the memory is allocated for top-level heaps and subheaps in shared pool. How can we know what are all the top-level heaps and subheaps and their hierarchy. Is it possible to get it from any X$ tables. Freeable chunks can be flushed out?If yes ,then why there should be another type as recreatable chunks? Please explain. Thanks in advance Syed |
Title: Message
- Re: (long) Top level heaps/subheaps Tanel Poder
- RE: (long) Top level heaps/subheaps Guerra, Abraham J
- RE: (long) Top level heaps/subheaps Jared Still
- Re: (long) Top level heaps/subheaps Mladen Gogala
- Re: (long) Top level heaps/subheaps Jared Still
- Re: (long) Top level heaps/subheaps Wolfgang Breitling
- Re: (long) Top level heaps/subheaps Jared Still