종이컵통신

거시기..?

주난군 — Tue, 05 May 2020 23:28:00 +0900

일상

끄적

듣다

공지는 이곳에....

null column은 data size 측정이 어떻게 되는가...

주난군 — Tue, 22 May 2012 12:04:30 +0900

급 모델러 분이 column에서 null 값이 있을때 용량산정을 물어보심..

익히 알고 있는바는 있었나 검증되지 않아 이번에 검증하도록 함미당. +_+)

다음과 같은 table을 생성한다.

SQL> desc null_table

Name Null? Type
----------------------- -------- ------------
COL1 VARCHAR2(1)
COL2 NUMBER(5)
COL3 CHAR(2)
COL4 VARCHAR2(10)
COL5 VARCHAR2(10)
COL6 VARCHAR2(10)

해당 object를 확인해보쟈-

SQL> select header_file, header_block
from dba_segments where segment_name='NULL_TABLE';
no rows selected

SQL> -- 11g 부터는 데이터 입력값이 있어야 segment 생성

데이터를 넣어보겠슴미당.

insert into null_table values(1,null,null,null,null,null);
insert into null_table values(1,1,null,null,null,null);
insert into null_table values(1,null,null,null,null,1);
insert into null_table values(1,null,null,1,null,1);
insert into null_table values(1,null,null,null,null,1);
insert into null_table values(null,null,null,null,null,null);
commit;

다시한번 슥삭- 해보면

SQL> select header_file, header_block
from dba_segments where segment_name='NULL_TABLE';

HEADER_FILE HEADER_BLOCK
----------- ------------
4 50682

쨔잔~ 하고 나옴-!
기타 등등을 알아보자-

SQL> select rowid, dbms_rowid.rowid_relative_fno(rowid) fno
2 dbms_rowid.rowid_block_number(rowid) blkno,
3 dbms_rowid.rowid_row_number(rowid) rowno
4 from null_table;

ROWID FNO BLKNO ROWNO
------------------ ---------- ---------- ----------
AAAS5YAAEAAAMX9AAA 4 50685 0
AAAS5YAAEAAAMX9AAB 4 50685 1
AAAS5YAAEAAAMX9AAC 4 50685 2
AAAS5YAAEAAAMX9AAD 4 50685 3
AAAS5YAAEAAAMX9AAE 4 50685 4
AAAS5YAAEAAAMX9AAF 4 50685 5

SQL> select block_size from v$datafile where file#=4;

BLOCK_SIZE
----------
8192

음... block size는 8k인걸 확인할수 있다.
check point를 따시~ 먹이고 datafile dump를 생성하도록 하쟈-

SQL> alter system checkpoint;

System altered.

SQL> alter system dump datafile 4 block 50685;

System altered.

userdump_dest 항목을 참조..

*** 2012-05-22 11:55:42.437

Block header dump: 0x0100c5fd

Object id on Block? Y

seg/obj: 0x12e58 csc: 0x00.1a4dcf itc: 2 flg: E typ: 1 - DATA

brn: 0 bdba: 0x100c5f8 ver: 0x01 opc: 0

inc: 0 exflg: 0

Itl Xid Uba Flag Lck Scn/Fsc

0x01 0x0004.012.000005b5 0x00c02a5c.011a.0c --U- 6 fsc 0x0000.001a4dd0

0x02 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000

bdba: 0x0100c5fd

data_block_dump,data header at 0x7ffc5adab064

===============

tsiz: 0x1f98

hsiz: 0x1e

pbl: 0x7ffc5adab064

76543210

flag=--------

ntab=1

nrow=6

frre=-1

fsbo=0x1e

fseo=0x1f66

avsp=0x1f3d

tosp=0x1f3d

0xe:pti[0] nrow=6 offs=0

0x12:pri[0] offs=0x1f93

0x14:pri[1] offs=0x1f8b

0x16:pri[2] offs=0x1f80

0x18:pri[3] offs=0x1f74

0x1a:pri[4] offs=0x1f69

0x1c:pri[5] offs=0x1f66

block_row_dump:

tab 0, row 0, @0x1f93

tl: 5 fb: --H-FL-- lb: 0x1 cc: 1

col 0: [ 1] 31

tab 0, row 1, @0x1f8b

tl: 8 fb: --H-FL-- lb: 0x1 cc: 2

col 0: [ 1] 31

col 1: [ 2] c1 02

tab 0, row 2, @0x1f80

tl: 11 fb: --H-FL-- lb: 0x1 cc: 6

col 0: [ 1] 31

col 1: *NULL*

col 2: *NULL*

col 3: *NULL*

col 4: *NULL*

col 5: [ 1] 31

tab 0, row 3, @0x1f74

tl: 12 fb: --H-FL-- lb: 0x1 cc: 6

col 0: [ 1] 31

col 1: *NULL*

col 2: *NULL*

tl: 5 fb: --H-FL-- lb: 0x1 cc: 1

col 0: [ 1] 31

tab 0, row 1, @0x1f8b

tl: 8 fb: --H-FL-- lb: 0x1 cc: 2

col 0: [ 1] 31

col 1: [ 2] c1 02

tab 0, row 2, @0x1f80

tl: 11 fb: --H-FL-- lb: 0x1 cc: 6

col 0: [ 1] 31

col 1: *NULL*

col 2: *NULL*

col 3: *NULL*

col 4: *NULL*

col 5: [ 1] 31

tab 0, row 3, @0x1f74

tl: 12 fb: --H-FL-- lb: 0x1 cc: 6

col 0: [ 1] 31

col 1: *NULL*

col 2: *NULL*

col 3: [ 1] 31

col 4: *NULL*

col 5: [ 1] 31

tab 0, row 4, @0x1f69

tl: 11 fb: --H-FL-- lb: 0x1 cc: 6

col 0: [ 1] 31

col 1: *NULL*

col 2: *NULL*

col 3: *NULL*

col 4: *NULL*

col 5: [ 1] 31

tab 0, row 5, @0x1f66

tl: 3 fb: --H-FL-- lb: 0x1 cc: 0

end_of_block_dump

음.. 다음과 같이 나왔다..
하나 하나 살펴보도록 하여욤..!

SQL> select * from null_table;

C COL2 CO COL4 COL5 COL6
- ---------- -- ---------- ---------- ----------
1

tab 0, row 0, @0x1f93
tl: 5 fb: --H-FL-- lb: 0x1 cc: 1
col 0: [ 1] 31
보다시피 첫번째 행에는 1 값밖에 보이지 않는다.
col 0 첫번째컬럼에 [ 1] 1바이트를 차지하고 있으며 31 16진수로 표현된 ascii code값임을 확인
NULL 값은 header정보에 포함이 되어있지 않음을 확인할수 있다.
더불어 cc 값에 주목하도록 하자-!

C COL2 CO COL4 COL5 COL6
- ---------- -- ---------- ---------- ----------
1 1

tab 0, row 1, @0x1f8b
tl: 8 fb: --H-FL-- lb: 0x1 cc: 2
col 0: [ 1] 31
col 1: [ 2] c1 02
마찬가지이다..
다만 두번째행은 column type이 number 이므로 31이 아닌 c1으로 표시되었다.

C COL2 CO COL4 COL5 COL6
- ---------- -- ---------- ---------- ----------
1 1

tab 0, row 2, @0x1f80
tl: 11 fb: --H-FL-- lb: 0x1 cc: 6
col 0: [ 1] 31col 1: *NULL*
col 2: *NULL*
col 3: *NULL*
col 4: *NULL*
col 5: [ 1] 31
여기서부터가 좀 흥미로운데 중간에 NULL에 대한 표시는 있으나
해당 byte에 대한 점유는하지 않은것으로 확인된다.
여전히 col6에 대한 정보는 표시 되지 않는다.

C COL2 CO COL4 COL5 COL6
- ---------- -- ---------- ---------- ----------
1 1 1

tab 0, row 3, @0x1f74
tl: 12 fb: --H-FL-- lb: 0x1 cc: 6
col 0: [ 1] 31
col 1: *NULL*
col 2: *NULL*
col 3: [ 1] 31
col 4: *NULL*
col 5: [ 1] 31
같은 패턴을 보이고 있는것이 확인된다.

C COL2 CO COL4 COL5 COL6
- ---------- -- ---------- ---------- ----------
1 1

tab 0, row 4, @0x1f69
tl: 11 fb: --H-FL-- lb: 0x1 cc: 6
col 0: [ 1] 31
col 1: *NULL*
col 2: *NULL*
col 3: *NULL*
col 4: *NULL*
col 5: [ 1] 31
위와 같다.

C COL2 CO COL4 COL5 COL6
- ---------- -- ---------- ---------- ----------

tab 0, row 5, @0x1f66
tl: 3 fb: --H-FL-- lb: 0x1 cc: 0
전부다 NULL 경우에는 아무것도 표시되지 않음을 알수있다..

그러므로 modeling 단계부터 data에 대한 분포도값을 정확히 측정하여
null column 이 많은경우 해당 column을 최대한 뒤로 배치하여 앞에서부터읽을수 있도록 조치하면되겠당.. 중간에 null 이 아닌 다른값이 들어와 있으면해당 부분까지 scan을 하는것도 비용처리가 될꺼라 생각이 된다..

하다보니 index에 대한 궁금증이 생겨서 테스트를 더 진행해 보았다

index를 생성 6번째 컬럼을 기준으로 index를 생성한다

SQL> create index COL6_NULL on NULL_TABLE(COL6);

SQL> select object_name ,object_id from dba_objects where object_name ='COL6_NULL';

OBJECT_NAME OBJECT_ID

----------- ----------

COL6_NULL 77401

tree level dump를 뜹니당-

SQL> alter session set events 'immediate trace name treedump level 77401';

Session altered.

가뿐하게 trace file 생성후 열어보도록 합니다

[oracle@jhbae trace]$ vi DB11G_ora_31214_COL6_NULL.trc

Trace file /u01/app/oracle/diag/rdbms/db11g/DB11G/trace/DB11G_ora_31214_COL6_NULL.trc

Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

With the Partitioning, OLAP, Data Mining and Real Application Testing options

ORACLE_HOME = /u01/app/oracle/product/11.2.0/dbhome_1

System name: Linux

Node name: jhbae

Release: 2.6.32-300.3.1.el6uek.x86_64

Version: #1 SMP Fri Dec 9 18:57:35 EST 2011

Machine: x86_64

Instance name: DB11G

Redo thread mounted by this instance: 1

Oracle process number: 27

Unix process pid: 31214, image: oracle@jhbae (TNS V1-V3)

*** 2012-05-22 14:06:46.957

*** SESSION ID:(35.215) 2012-05-22 14:06:46.957

*** CLIENT ID:() 2012-05-22 14:06:46.957

*** SERVICE NAME:(SYS$USERS) 2012-05-22 14:06:46.957

*** MODULE NAME:(SQL*Plus) 2012-05-22 14:06:46.957

*** ACTION NAME:() 2012-05-22 14:06:46.957

----- begin tree dump

leaf: 0x100c603 16827907 (0: nrow: 3 rrow: 3)

----- end tree dump

도출된 값을 가지고 dump file을 생성합니다.

[oracle@jhbae trace]$ sqlplus / as sysdba

SQL*Plus: Release 11.2.0.3.0 Production on Tue May 22 14:08:00 2012

Copyright (c) 1982, 2011, Oracle. All rights reserved.

Connected to:

Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

With the Partitioning, OLAP, Data Mining and Real Application Testing options

SQL> select dbms_utility.data_block_address_file(16827907) as file_no,

2 dbms_utility.data_block_address_block(16827907) as block_no from dual;

FILE_NO BLOCK_NO

---------- ----------

4 50691

Elapsed: 00:00:00.05

SQL> alter system dump datafile 4 block 50691;

System altered.

Elapsed: 00:00:00.03

[oracle@jhbae trace]$ vi DB11G_ora_32179.trc

Trace file /u01/app/oracle/diag/rdbms/db11g/DB11G/trace/DB11G_ora_32179.trc

Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

With the Partitioning, OLAP, Data Mining and Real Application Testing options

ORACLE_HOME = /u01/app/oracle/product/11.2.0/dbhome_1

System name: Linux

Node name: jhbae

Release: 2.6.32-300.3.1.el6uek.x86_64

Version: #1 SMP Fri Dec 9 18:57:35 EST 2011

Machine: x86_64

Instance name: DB11G

Redo thread mounted by this instance: 1

Oracle process number: 27

Unix process pid: 32179, image: oracle@jhbae (TNS V1-V3)

*** 2012-05-22 14:25:19.217

*** SESSION ID:(35.221) 2012-05-22 14:25:19.217

*** CLIENT ID:() 2012-05-22 14:25:19.217

*** SERVICE NAME:(SYS$USERS) 2012-05-22 14:25:19.217

*** MODULE NAME:(sqlplus@jhbae (TNS V1-V3)) 2012-05-22 14:25:19.217

*** ACTION NAME:() 2012-05-22 14:25:19.217

Start dump data blocks tsn: 4 file#:4 minblk 50691 maxblk 50691

Block dump from cache:

Dump of buffer cache at level 4 for tsn=4 rdba=16827907

BH (0x71bdbca8) file#: 4 rdba: 0x0100c603 (4/50691) class: 1 ba: 0x71868000

set: 3 pool: 3 bsz: 8192 bsi: 0 sflg: 1 pwc: 397,28

dbwrid: 0 obj: 77401 objn: 77401 tsn: 4 afn: 4 hint: f

hash: [0x922132e0,0x922132e0] lru: [0x71bdbec0,0x71bf4ef0]

ckptq: [NULL] fileq: [NULL] objq: [0x643fc118,0x88732b38] objaq: [0x6a7db578,0x88732b28]

st: XCURRENT md: NULL fpin: 'ktspbwh2: ktspfmdb' tch: 2

flags: block_written_once redo_since_read

LRBA: [0x0.0.0] LSCN: [0x0.0] HSCN: [0xffff.ffffffff] HSUB: [1]

Block dump from disk:

buffer tsn: 4 rdba: 0x0100c603 (4/50691)

scn: 0x0000.001a66e5 seq: 0x01 flg: 0x04 tail: 0x66e50601

frmt: 0x02 chkval: 0xb9e8 type: 0x06=trans data

Hex dump of block: st=0, typ_found=1

seg/obj: 0x12e59 csc: 0x00.1a66e3 itc: 2 flg: E typ: 2 - INDEX

brn: 0 bdba: 0x100c600 ver: 0x01 opc: 0

inc: 0 exflg: 0

Itl Xid Uba Flag Lck Scn/Fsc

0x01 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000

0x02 0xffff.000.00000000 0x00000000.0000.00 C--- 0 scn 0x0000.001a66e3

Leaf block dump

===============

header address 139870805363300=0x7f3635ab0a64

kdxcolev 0

KDXCOLEV Flags = - - -

kdxcolok 0

kdxcoopc 0x80: opcode=0: iot flags=--- is converted=Y

kdxconco 2

kdxcosdc 0

kdxconro 3

kdxcofbo 42=0x2a

kdxcofeo 7999=0x1f3f

kdxcoavs 7957

kdxlespl 0

kdxlende 0

kdxlenxt 0=0x0

kdxleprv 0=0x0

kdxledsz 0

kdxlebksz 8032

row#0[8021] flag: ------, lock: 0, len=11

col 0; len 1; (1): 31

col 1; len 6; (6): 01 00 c5 fd 00 02

row#1[8010] flag: ------, lock: 0, len=11

col 0; len 1; (1): 31

col 1; len 6; (6): 01 00 c5 fd 00 03

row#2[7999] flag: ------, lock: 0, len=11

col 0; len 1; (1): 31

col 1; len 6; (6): 01 00 c5 fd 00 04

----- end of leaf block dump -----

위에서 도출된값과 같이 len (1) 16진수로 31을 가리키고 있는것을 확인할수 있음.
그렇다면 두번째 커럼인 col 1의 정체는 무엇인가... 그것은 바로..!
row id 값... 그것을 확인해 봅니당-

SQL*Plus: Release 11.2.0.3.0 Production on Tue May 22 14:54:11 2012
Copyright (c) 1982, 2011, Oracle. All rights reserved.

Connected to:

Oracle Database 11g Enterprise Edition Release 11.2.0.3.0 - 64bit Production

With the Partitioning, OLAP, Data Mining and Real Application Testing options

SQL> select rowid extended_format,
dbms_rowid.rowid_to_restricted(rowid,0),col1,col2,col3,col4,col5,col6
2 from null_table;

EXTENDED_FORMAT DBMS_ROWID.ROWID_T C COL2 CO COL4 COL5 COL6

------------------ ------------------ - ---------- -- ---------- ---------- ----------

AAAS5YAAEAAAMX9AAA 0000C5FD.0000.0004 1

AAAS5YAAEAAAMX9AAB 0000C5FD.0001.0004 1 1

AAAS5YAAEAAAMX9AAC 0000C5FD.0002.0004 1 1

AAAS5YAAEAAAMX9AAD 0000C5FD.0003.0004 1 1 1

AAAS5YAAEAAAMX9AAE 0000C5FD.0004.0004 1 1

AAAS5YAAEAAAMX9AAF 0000C5FD.0005.0004

6 rows selected.

Elapsed: 00:00:00.01

index는 null값을 저장하지 않는것으로 확인하였습니다-
그리고 rowid는 위와같이 변환해서 보면 rowid_to_restricted 부분과

index dump file의 col1 이 일치되는것을 확인할수 있음!

끗!

'6. 먹고살기 > 6.1 Oracle' 카테고리의 다른 글

null column은 data size 측정이 어떻게 되는가... (0)	12:04:30
Hadoop and NoSQL Technologies and the Oracle Database (0)	2012/05/21
object statistics are locked (0)	2012/03/30
COMMIT_POINT_STRENGTH (0)	2012/01/19

Hadoop and NoSQL Technologies and the Oracle Database

주난군 — Mon, 21 May 2012 17:14:36 +0900

An Oracle White Paper

February 2011

Hadoop and NoSQL Technologies and the Oracle Database

Disclaimer

The following is intended to outline our general product direction. It is intended for information purposes only, and may not be incorporated into any contract. It is not a commitment to deliver any material, code, or functionality, and should not be relied upon in making purchasing decisions. The development, release, and timing of any features or functionality described for Oracle‟s products remains at the sole discretion of Oracle.

Hadoop and NoSQL Technologies and the Oracle Database

Introduction

Innovation in the database software industry continues to be extremely vibrant. In just the last couple of years, new scalability capabilities have revolutionized the performance that database systems can deliver. These include the use of very large, multi-terabyte flash memories; compression which can dramatically increase the amount of data cached in memory and the speed of disk scans for large queries; and the movement of select database logic into storage to speed functions such as row selection, compression, encryption, and more. For Oracle, this has culminated in Exadata which is an integrated software and hardware product incorporating all of these new innovations in a scale-out architecture that effectively provides unlimited scalability for all types of applications.

The world of open source database technology has also been evolving rapidly. In this paper we will focus on two relatively new developments: Hadoop (MapReduce) and NoSQL. The common goal for both is massive scalability and support for what is called “Big Data”, i.e., environments where data can range from tens or hundreds of terabytes to petabytes or more in size. The purpose of this paper is to explain the use cases for these technologies and how they relate to and coexist with the Oracle Database.

Open Source Database Technologies

Open source Database technologies have been used to build and successfully run some of the most innovative new applications that have emerged in recent years. Developers with new ideas have been able to quickly develop and then deploy these applications at a low cost with open source databases. They didn‟t need to spend time and energy convincing others to invest in the purchase of commercial database products or hire the expertise of professional DBAs.

Today, Oracle Corporation is the leading open source database vendor. Oracle acquired InnoDB, the leading transactional storage engine for MySQL, in 2005. Berkeley DB, a high-performance embedded database for key/value data, was acquired in 2006. MySQL, the leading open source SQL-based relational database, was acquired in 2010. These products have thrived since their acquisition and Oracle has continued to invest significantly in their development.

Many Oracle customers have substantial deployments of both the Oracle Database and open source databases. Many of the new, innovative applications that use open source databases today require a relatively narrow range of functionality when compared to more mainstream business applications for which the Oracle Database is by far the leading platform. Over time some of these applications will need to provide more complex

Hadoop and NoSQL Technologies and the Oracle Database

functionality to remain competitive in their market and will need to move to or integrate with the database platform that can best support these additional requirements. Oracle has been working to support integration between these environments where that can add value. Oracle has also been working to support low-cost and low-risk migrations from open source databases to the Oracle Database when this is the next step in the application lifecycle.

Hadoop

Hadoop is a generic processing framework designed to execute queries and other batch read operations against massive datasets that can be tens or hundreds of terabytes and even petabytes in size. The data is loaded into or appended to the Hadoop Distributed File System (HDFS). Hadoop then performs brute force scans through the data to produce results that are output into other files. It probably does not qualify as a database since it does not perform updates or any transactional processing. Hadoop also does not support such basic functions as indexing or a SQL interface, although there are additional open source projects underway to add these capabilities.

Hadoop operates on massive datasets by horizontally scaling (aka scaling out) the processing across very large numbers of servers through an approach called MapReduce. Vertical scaling (aka scaling up), i.e., running on the most powerful single server available, is both very expensive and limiting. There is no single server available today or in the foreseeable future that has the necessary power to process so much data in a timely manner.

Hundreds or thousands of small, inexpensive, commodity servers do have the power if the processing can be horizontally scaled and executed in parallel. Using the MapReduce approach, Hadoop splits up a problem, sends the sub-problems to different servers, and lets each server solve its sub-problem in parallel. It then merges all the sub-problem solutions together and writes out the solution into files which may in turn be used as inputs into additional MapReduce steps.

Hadoop has been particularly useful in environments where massive server farms are being used to collect the data. Hadoop is able to process parallel queries as big, background batch jobs on the same server farm. This saves the user from having to acquire additional hardware for a database system to process the data. Most importantly, it also saves the user from having to load the data into another system. The huge amount of data that needs to be loaded can make this impractical.

Hadoop and NoSQL Technologies and the Oracle Database

Hadoop Use Cases

Many of the ideas behind the open source Hadoop project originated from the Internet search community, most notably Google and Yahoo. Search engines employ massive farms of inexpensive servers that crawl the web retrieving web pages into local files. They then process this data using massive parallel queries to build indexes to enable search.

The actual algorithms used by search engines to determine the relevance to search terms and quality of web pages are very specialized, sophisticated, and highly proprietary. These algorithms are the secret sauce of search. The application of the algorithms to each webpage and the aggregation of the results into indexes to enable search is done through MapReduce processes and is more straightforward although done on a massive scale. The map function identifies each use of a potential search parameter in a webpage. The reduce function aggregates this data, e.g., determines the number of times a search parameter is used in a page.

There are use cases for Hadoop in many other industries. Some large websites use Hadoop to analyze usage patterns from log files or click-stream data that is generated by hundreds or thousands of their webservers. The scientific community can use Hadoop on huge server farms that monitor natural phenomena and/or the results of experiments. The intelligence community needs to analyze vast amounts of data gathered by server farms monitoring phone, email, instant messaging, travel, shipping, etc. to identify potential terrorist threats. The publishing industry may use Hadoop to index and/or reformat huge document stores. It is important to note, however, that not all highly parallel tasks are amenable to efficient MapReduce solutions.

Hadoop and the Oracle Database

Sometimes the processing needed is just a one-off. For example, a search company needs to use the search indexes generated from today‟s web contents. Tomorrow those indexes will be replaced by processing tomorrow‟s data. The processing might also be self contained. Publishing data might never need to be combined with other data the publisher may possess or acquire. In some cases, though, more complex processing of historical data with current data or with entirely different types of data will be needed that require a different type of system.

Many open source and commercial database products do not support operations such as parallel query through horizontal scaling but there are a number that do. Some of the commercial database products in particular are very advanced and offer many, many capabilities that Hadoop does not. These include many very powerful performance optimizations, sophisticated analytic functions, and ease of use features, including rich declarative features that allow even very complex analysis to be done by non-programmers. These also include enterprise class features for security, auditing, maximum availability, disaster recovery, etc. These capabilities have been produced and refined over decades of development in an extremely competitive marketplace. The Oracle Database is both the technology leader and the market share leader in this area. Again, Exadata is

Hadoop and NoSQL Technologies and the Oracle Database

Oracle‟s most recent, major leap in new capability for these and other types of applications.

When richer capabilities are needed, commercial databases like Oracle may not replace Hadoop but rather coexist with it. Hadoop and the Oracle Database complement each other. The sheer magnitude of data involved makes it very sensible to use the “cheap cycles” of server farms to transform masses of unstructured data with low information density into smaller amounts of information dense structured data that is then loaded into Oracle. Advanced mechanisms for loading the data that make maximum use of both Hadoop and Oracle capabilities would also be very useful.

NoSQL

There are an extremely large number of NoSQL databases being developed. Some of the more prominent include Apache Cassandra, MongoDB, Voldemort, Apache HBase, SimpleDB, and BigTable. New ones seem to pop up regularly, gain prominence, and then fade from favor. All of these technologies are in the early stages of development and most of the early adopters employ programming staffs that participate in their development. The Oracle open source database Berkeley DB is used as a storage engine for Dynamo, Voldemort, GenieDB and others. Oracle InnoDB is used as a storage engine for Voldemort, Riak, and others.

Some advocates of the term “NoSQL” emphasize that this is a distinctly different approach from SQL-based databases and is fundamentally non-relational. Others soften the term by saying “NoSQL” stands for “Not Only SQL”. Indeed, there are projects underway to build SQL interfaces to NoSQL databases.

The focus of this paper is on NoSQL databases that are designed to execute very large volumes of simple updates and reads against a single, very large dataset. They are designed to handle the processing volume needed to support millions and potentially even hundreds of millions of online users. The datasets involved, although typically not as large as Hadoop is targeted at, may reach tens or hundreds of terabytes in size or larger.

Like Hadoop, these NoSQL databases do this by horizontally scaling across very large numbers of servers, e.g., tens or hundreds or even thousands of servers. The technique used by these NoSQL databases for horizontal scaling is essentially the same as has been deployed for many years to support very high volume systems using conventional relational databases. The technique is called sharding. It requires that a separate database run on each server and that the data be physically partitioned so that each database has its own subset of the data stored on its own local disks.

Hadoop and NoSQL Technologies and the Oracle Database

A problem with sharding, however, is that you lose key relational database capabilities when you shard. Specifically, you lose much of your ability to do joins, transactions, and schema changes. You also need to compromise ACID (atomicity, consistency, isolation, and durability) principles, usually consistency must be relaxed.

The core idea behind the NoSQL databases is that if you are going to lose all this capability when you shard why complicate the database by supporting these features. Instead, support other features that provide some partial functionality in some of these areas and, most of all, try to simplify and automate what is the biggest challenge - the sharding mechanism itself.

Lost Capabilities

To join any data with any other data in a sharded environment you need to be able to do distributed queries that span potentially very large numbers of separate databases. Distributed queries are more complicated than queries that join data within a single database. These complications can create overhead and be prohibitively slow. NoSQL databases, therefore, typically don‟t support joins.

Similarly, transactions that allow updates to multiple rows to be committed together (or all rolled back should a failure occur) require distributed transactions in a sharded environment. Distributed transactions require some form of two phase commit protocol which is more complicated and expensive than what is needed if all the data is contained within a single database Two phase commit can be slow, and worse, can compromise availability because failures can cause data to become locked and inaccessible (called „indoubt transactions”) until the failure is repaired. NoSQL databases, therefore, typically don‟t support transactions that involve updates to data in multiple tables or even multiple rows within a table.

It is also prohibitively difficult to make schema changes in a sharded environment. Some databases enable a schema change, such as adding a column to a table, to be done while the system is online within one database, but not across multiple databases. If you attempt to make a schema change across each database individually then the application will sometimes see the column in the table and sometimes not. If the application attempts to access the column but it is not there it will fail. NoSQL databases, therefore, typically don‟t support online schema changes. Many are even “schema-less”. The application becomes entirely responsible for keeping track of the format of the data in the database and dealing with the different formats that it may see.

Partial Functionality

To provide partial functionality is some of these areas, NoSQL databases typically support some mechanism for adding new data on the fly. There is a great deal of variation in the exact functionality that the many different NoSQL databases provide (including document, columnar, and key/value approaches) but the core idea is that there be some way to add new data elements or columns to tables and the values contained in those

Hadoop and NoSQL Technologies and the Oracle Database

columns in a manner that applications can readily detect when the new columns are there or not.

Key/value support is one flexible and straightforward way of doing this. Key/values make it possible to add additional logic into applications to deal with online schema changes. Every data value is accompanied by a key which labels the data much like a column name. Applications can be programmed to test and appropriately deal with whether a new column and its value exist yet or not on a row by row basis.

Key/values also provide partial functionality to deal with the lack of joins by allowing de-normalized or pre-joined data to be stored in a single data set (or table). For example, say you need to store data for a blogging system where you have blog entries and comments on blog entries. With a relational system you could store them in a normalized format with blog entries in one table and comments in another with the appropriate foreign key links and join them when you need to. With key/value pairs you would have a dataset (table) containing the blog entries and then add an arbitrary list of key/value pairs to each row to store the comments in each blog entry.

Sharding

Simplifying and automating sharding is probably the most difficult challenge facing the NoSQL databases. A way needs to be found to distribute or partition the data across potentially huge numbers of servers in a manner that allows each data item to be readily located when needed and, very importantly, to balance the processing load evenly across all the servers. The biggest challenge is handling changes that force you to rebalance the distribution of the data, or repartition. This can be very difficult on a huge system with vast amounts of data and vast numbers of users always online. It is not clear whether any of the current NoSQL products have solved all these problems.

Manual sharding schemes are typically highly application dependant and require customized solutions with repeated revisions to make them work. For example, a social network may choose to partition based on user name. All users whose last name starts with “A” on one server, “B” on another, and so on. It would need to be more complicated than this since there are probably far more people whose name starts with an “A” in the world than “Z” but this can be worked out at least initially using various techniques such as lookup tables and hashing. What happens, though, when your service becomes suddenly popular in a country where people‟s last names do frequently start with “Z”? The data will need to be repartitioned.

Producing a generic auto-sharding solution that really works in production is difficult. There have been a number of attempts at using various forms of consistent hashing as the basis for auto-sharding. Some keys, though, have inherent problems with even distributions such as states or countries, e.g., a social network will probably have many more members from California than Rhode Island. Unique keys, like user account id,

Hadoop and NoSQL Technologies and the Oracle Database

often have balanced load distributions when hashed but have other challenges such as dealing with changes in overall processing load.

For example, what happens if you are running on a thousand servers and discover that the overall system needs 10% more processing power, or 100 additional servers, to handle the load? Cassandra, one of the newer NoSQL technologies, would add this additional capacity by identifying the 100 most heavily used servers and splitting the data on each server with one of the new servers. This may not work well:

 If the load was unbalanced so that there were 100 servers that were seriously overloaded then splitting those 100 servers may be the last thing you want to do. Splitting a server requires that you copy half of its data to the new server. That is a big operation to perform when you are already overloaded. Access to data on these servers may become unacceptably slow. Still worse, very heavy load conditions can cause systems to fail in many unexpected ways that could cascade to other systems, and so on.

 If the load was evenly balanced across the 1000 servers adding more servers probably won‟t do what you want either. 100 of these servers would now have double the capacity to handle their load and, in this case, about 90% of the additional capacity would go unused, while the other 900 servers would continue to be overloaded.

NoSQL and the Oracle Database

NoSQL products do have the advantage that because they are open source they may be more accessible to developers building entirely new, highly innovative applications. The applications, though, must be simple enough to fit within the constraints and limitations that NoSQL databases inherently possess.

When applications become more complex the Oracle Database provides a very natural next step in the application lifecyle because the Oracle Database supports the data and processing volumes being targeted by the NoSQL database. It does this by horizontally scaling across both servers and storage with Oracle technologies such as Real Application Clusters (RAC) and Automatic Storage Management (ASM) without the restrictions of the NoSQL approach.

With Real Application Clusters running on the database servers and Automatic Storage Management managing the storage you can horizontally scale across very large numbers of servers and storage devices but it is all one database. These technologies do not need or use distributed queries or two-phase commit to support joins or transactions. Also, schema changes can be done online which greatly simplifies application development.

Sometimes it does make sense for performance reasons to store data in de-normalized ways if the application is always going to access child table data through the parent table.

Hadoop and NoSQL Technologies and the Oracle Database

For example, you may know that comments on a blog entry are only going to be accessed when you are reading the blog entry. Oracle, though, has mechanisms for this, such as clustered indexes, that allow you to store the data in a pre-joined format for maximum performance.

Oracle Database technology is also extremely effective at dynamically balancing processing load across database servers and storage. Its forte is allowing for change. Servers and storage can be added or removed online. All data is accessible to all servers all the time. The processing involved in adding servers and storage is evenly distributed across the whole system so that no one component is over loaded.

Most of all, this technology is proven. Thousands of the largest and most mission critical systems in the world today have been running on the Oracle Database with RAC and ASM technologies with great success for years.

Conclusion

Open source Database technologies have been used to build and successfully run some of the most innovative new applications that have emerged in recent years. This is primarily because the developers of these applications have found open source databases to be very accessible. Developers with radical new ideas have been able to develop and deploy these applications quickly without the need to spend time and energy convincing others to invest in the purchase of commercial database products or the expertise of professional DBAs. Oracle today is the largest open source database vendor, with InnoDB, Berkeley DB, and MySQL to address this market need.

Many Oracle Database customers also have substantial deployments of open source databases. Oracle, therefore, has been working to support integration between these environments where that can add value. Hadoop and the Oracle Database complement each other in environments where massive server farms collect data that needs sophisticated analysis.

Oracle has also been working to support low-cost and low-risk migrations from open source databases to the Oracle Database. Many of the new, innovative applications that use open source databases today will need to provide more complex functionality to remain competitive in their market in the future and will need to move to the database platform that can best support that. The best, and really only solution, for complex applications that require extreme scalability and support for “Big Data” is the Oracle Database.

Hadoop and NoSQL Technologies and the Oracle Database January 2011 Author: Gordon Smith

Oracle Corporation World Headquarters 500 Oracle Parkway Redwood Shores, CA 94065 U.S.A.

Worldwide Inquiries: Phone: +1.650.506.7000 Fax: +1.650.506.7200

oracle.com

Copyright © 2011, Oracle and/or its affiliates. All rights reserved. This document is provided for information purposes only and the contents hereof are subject to change without notice. This document is not warranted to be error-free, nor subject to any other warranties or conditions, whether expressed orally or implied in law, including implied warranties and conditions of merchantability or fitness for a particular purpose. We specifically disclaim any liability with respect to this document and no contractual obligations are formed either directly or indirectly by this document. This document may not be reproduced or transmitted in any form or by any means, electronic or mechanical, for any purpose, without our prior written permission.

Oracle and Java are registered trademarks of Oracle and/or its affiliates. Other names may be trademarks of their respective owners.

AMD, Opteron, the AMD logo, and the AMD Opteron logo are trademarks or registered trademarks of Advanced Micro Devices. Intel and Intel Xeon are trademarks or registered trademarks of Intel Corporation. All SPARC trademarks are used under license and are trademarks or registered trademarks of SPARC International, Inc. UNIX is a registered trademark licensed through X/Open Company, Ltd. 1010

'6. 먹고살기 > 6.1 Oracle' 카테고리의 다른 글

null column은 data size 측정이 어떻게 되는가... (0)	12:04:30
Hadoop and NoSQL Technologies and the Oracle Database (0)	2012/05/21
object statistics are locked (0)	2012/03/30
COMMIT_POINT_STRENGTH (0)	2012/01/19

밤에는...

주난군 — Sat, 19 May 2012 00:56:11 +0900

오랜만에 글을 쓰려고 생각하기 전에는 말이다. 특히 이렇게 목적도 없이 아무생각도 없이 글을쓰려고 뭔가. 음.. 하고 하기 전에는. 굳이 일기장을 볼필요가 없다. 그러니깐 군대에서 썼던 일기장 같은것 말이다. 사람이라면 누구든지 그러한 시기를 거친다.. 애타고 바둥대고 환경은 절제 되어 있는데 혼자만 홀로 불탈때.

그리고 그게 조금은 잘못된 방향으로 하염없이 가고 있는데 나는 정말 무기력한 시간들 말이다. 군대 이후로 무엇이 변하였는가 하면. 아무것도 변하지 않았다..라고도 할수 있고 많은것이 변했다.. 라고도 할수 있다.

나는 일단 정민기를 안다. ~~아 좋다. 좋은 일이다. 군대에서 얻은 가장 값비싼 선물이다.~~ 그는 다르게 변화하고 있었다. 어떠한 의미를 가지는지. 어떠한 방향성을 가지는지는 별개의 문제로. 그는 여전히 변하고 있었다. 이를테면 허브에서 라우터로 혹은 게이트웨이로 말이지. 좋다. 그는 좋다. 그의 어떤부분이 모가나고 욕을 먹는다 해도 그는 좋은거다. 내가 그를 받아들였고, 그 역시 나를 일부 받아들였음에 가능한 일이다. 좋다.

뱃살. 뱃살이 여전하다.
실은 더 쪘다.
살은. 왜 나를.

가만두지 않는걸까.

자고 있을때 와서 나를 괴롭히는걸까.. 왜..?
여전한것중에 하나다.

책. 나는 여전히 책을 읽는다. 군대에 있어서보단 양과 질이 많이 줄었지만. 생각하는것도 줄고 써내려가는 양도 줄었지만, 나는 여전히 읽는다. 읽는게 즐겁다. 나는 하나의 입과 두가지 귀중에서 고르라 하면, 세개의 귀와 입은 택하지 않을 생각이다. 여전히 읽고, 쓰는건 좋다. 내가 쓰는게 아니라. 나라는 어떠한 매개체로 인해 전혀 다른성질의 결과물이 나온다는게 가끔은 경이로울때가 있다. 전적으로 스스로만. 경이롭다.

박민규.
그는 왜 신작을 내지 않는가.
씁슬할 정도다. 나아~쁜 사람.

그때보단 조금 많이 벌고 있다. 그뿐이다.

그리고 옆에 나를 아주 많이 좋아해주는 사람이 있다. 이건 그뿐이 아니라. 너무 좋다.
그러니까 전적으로 너.무.좋.다. 라고 쓸수 밖에 없는건 어떤 거대 담론을 만났을때 으레 겁을 집어 먹거나
사람이 경직되는 성격과 같다. 나는 나를. 너.무.좋.아. 하는 사람을 만나고 있다.
항상 고맙고도 미안한 마음이다.

나이가 늘었다. 이제 내년이면 서른이다아- 광석형의 노래가 어울릴만한 나이가 되어감에-
행여-
나이에 맞지 않는 삶과 행동양식에 반성을 한다. 참.. 반성을 해도 나아지지 않음에 반성을 하고.
반성으로 끝나는 삶을 되돌아 반성한다.
매우 나쁜일이 아닐수 없다.

나쁘다.

이제는 잠을 자야겠다.
어디서 부터 풀어야할지 모르는 삶의 실타래를 너무도 많이 방치해두었던 시간이 나를 괴롭힌다.
자기자신을 사랑하지 않는 사람이.
무엇을 한다는것 자체가 이상한 일인거겠지.

어쩌라고....

'2.쓰다 > 2.1 일상' 카테고리의 다른 글

밤에는... (0)	2012/05/19
나쁜 티스토리.. (0)	2012/04/08
잠시.. (0)	2011/12/01
암튼.. (2)	2011/09/14

나들잉-

주난군 — Sun, 22 Apr 2012 20:19:54 +0900

배나온건 에러
입벌리고 찍은건 안 에러
졸린건 에러-

그러고 보니 만난지도 이제 10년...

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

나들잉- (0)	2012/04/22
땡이 (2)	2011/08/12
살아있어서 (0)	2011/06/29
@집 (0)	2011/01/28

주진우의 정통시사활극 주기자

주난군 — Mon, 09 Apr 2012 00:42:32 +0900

사람들은 말한다. 인생은 그런 게 아니라고. 강하면 부러진다고. 나도 편히 사는법을 안다. 좋은게 좋은 거라는 의미도 안다. 이러한 합리적인 이성은 실패에서 멀어지게 만든다. 동시에 나를 꿈에서도 떼어 놓으려고 한다. 나는 사랑하는 가슴으로 불가능한 꿈을 꾸면서 살겠다. 그 가슴은 영원히 상처받지 않고, 신의 보살핌을 받는다고 주문을 외우면서 이성을 넘어 가슴을 따르고 가슴으로 판단하겠다. 깨지고 부서지더라도 충동을 믿고 도전하겠다. 강자에게는 당당함으로, 약자에게는 겸손함으로 세상에 보탬이 되겠다. 이상과 정의 그리고 진실을 위해서는 그 어떤 타협도 하지 않겠다.. 에필로그

삼성이랑 싸운다. 교회랑 싸운다. 검경 대검 거칠것이 없다.
그러면서 부끄럽다 한다. 나는 삼성. 교회. 대검. 검경 부러운데.

부러우면 지는건데 분명 부럽다. 아직도 생각해야할것과 고민해야될것이.
20대에서 끝나지 않고 29 그리고 30대로 이어진다.

얼마전에 지인과 정치이야기를 했다. 참나원. 정치이야기는 왠만해선 안하는데..
그래서 말을 하다가 "그래서 넌 지지하는게 누군데?" 원참나.
그나마 가까운 심상정 생각이 나서 그녀를 이야기 했다.
그는 그사람이 싫다 하였다. 너무 데모를 많이 하고 정권심판론 어쩌구 이야기를 한다. 나원참.

근데 알고는 있으려나 모르겠다.
그런 사람들이 우릴위해 싸우는걸..

암튼 나도 요즘 투표근 단련중이다.
4월 11일날 투표해야지..

왜 시사활극인지는 본사람만이 안다 +_+)
프롤로그보단. 에필로그가 좋았당-!

'1.보다 > 1.1 책. 그리고, 밑줄' 카테고리의 다른 글

주진우의 정통시사활극 주기자 (2)	2012/04/09
책 지름- (0)	2012/04/08
책____ (0)	2011/04/20
@공씨책방 (0)	2011/02/18

책 지름-

주난군 — Sun, 08 Apr 2012 20:35:35 +0900

참.. 책샀땅-!

리춘식 형님 책이 있어서 늦게 올줄 알았더니..

하루만에 도착-! 열씨미 공부하겠슴니당.. 횽님 (__)

'1.보다 > 1.1 책. 그리고, 밑줄' 카테고리의 다른 글

주진우의 정통시사활극 주기자 (2)	2012/04/09
책 지름- (0)	2012/04/08
책____ (0)	2011/04/20
@공씨책방 (0)	2011/02/18

나쁜 티스토리..

주난군 — Sun, 08 Apr 2012 19:57:15 +0900

으.... 티스토리

자꼬 에디터 이상하게 만들지말구 신텍스 하이라이트 기능이나 넣어달라구..

slq문이 이상하게 나온단 말야..!

이게 모니 ㅠㅠㅠ

아 내가 뜯어 고쳐따 ㅠㅠ

SELECT SYS_CONTEXT('USERENV', 'SERVER_HOST'),
SYS_CONTEXT('USERENV', 'DB_UNIQUE_NAME'),
SYS_CONTEXT('USERENV', 'INSTANCE_NAME'),
SYS_CONTEXT('USERENV', 'SERVICE_NAME'),
INSTANCE_NUMBER,
STARTUP_TIME,
SYS_CONTEXT('USERENV', 'DB_DOMAIN')
FROM v$instance
WHERE INSTANCE_NAME=SYS_CONTEXT('USERENV', 'INSTANCE_NAME')

'2.쓰다 > 2.1 일상' 카테고리의 다른 글

밤에는... (0)	2012/05/19
나쁜 티스토리.. (0)	2012/04/08
잠시.. (0)	2011/12/01
암튼.. (2)	2011/09/14

object statistics are locked

주난군 — Fri, 30 Mar 2012 13:28:13 +0900

expdp 중 metadata_only 옵션을 이용하여 해당 테이블 정보를 뽑다가 이상한일이 생겨서 조치-

해당 schema에서 content=metadata_only 옵션을 이용하여 ddl문 추출

impdp 에서 해당 remap_schema, remap_tablespace, content=metadata_only 를 이용하여

새로운 계정생성 이상없이 끝나서 해당 계정을 select num_rows from user_tables 해보니 다음과 같은결과..

SQL> select num_rows from user_tables; 
  NUM_ROWS          
----------------- 
         0 
         5 
        14 
         4 
         7 
         7 
         2 
         4 
        10

아놔....

이 num_rows는 뭐지... count(*) 로 확인하면 될껄 굳이 num_rows로 확인 -_-;;

생각을 해보니.. 통계정보까지 같이 들어가는걸 깜박..

count하면 해당 row는 없다.. 그래두 깔끔하게 하려고 통계정보 생성- 암튼..

통계정보야 재 생성하면 되니깐...라는 생각으로 다음을 실행

SQL> exec dbms_stats.gather_table_stats('RECOTEST','EMP'); 

BEGIN dbms_stats.gather_table_stats('RECOTEST','EMP'); END; 



* 

ERROR at line 1: 

ORA-20005: object statistics are locked (stattype = ALL) 

ORA-06512: at "SYS.DBMS_STATS", line 23829 

ORA-06512: at "SYS.DBMS_STATS", line 23880 

ORA-06512: at line 1

ORA-2005 의 에러메시지가 나타나는 경우는 크게 3가지 인데 다음과 같다.

1. dbms_stat.lock[schema|table]_stat 를 이용해서 해당 통계정보를 잠글때

2. import 및 impdp를 사용하여 데이터 결과값 없이 table만 생성하는경우

3. queue table 인 경우 dynamic sampling이 사용되도록 통계정보가 비워지거나 잠기게 되며

3.1 10gR2로 upgrade 중에는 queue table 들의 통계정보는 삭제된 후 잠김상태가 되며10gR2 에서 queue table이 생성될때 통계들은 계속 비워진 채로 잠기게 된다.

~> 해결방법 imp사용시에는 rows=n, statistics=none 의 옵션을 사용하여 뽑도록하고 impdp 사용시는 content=metadata_only, exclude=(table_statistics,index_statistics) parameter를 사용하도록 한다.

일단 다시 뽑기는 귀찮으니 다음 테이블에서 통계정보가 잠긴 table을 확인한다.

SQL> select owner, table_name, stattype_locked  
from dba_tab_statistics where stattype_locked is not null; 


OWNER                          TABLE_NAME                     STATT 
------------------------------ ------------------------------ ----- 
SYS                            SCHEDULER$_EVENT_QTAB          ALL 
SYS                            SCHEDULER$_REMDB_JOBQTAB       ALL 
SYS                            SCHEDULER_FILEWATCHER_QT       ALL 
SYS                            ALERT_QT                       ALL 
SYS                            AQ$_MEM_MC                     ALL 
SYS                            AQ_EVENT_TABLE                 ALL 
SYS                            AQ_PROP_TABLE                  ALL 
SYS                            AQ$_KUPC$DATAPUMP_QUETAB_1_P   ALL 
SYS                            KUPC$DATAPUMP_QUETAB           ALL 
SYSTEM                         DEF$_AQCALL                    ALL 
SYSTEM                         DEF$_AQERROR                   ALL 
SYS                            SYS$SERVICE_METRICS_TAB        ALL 
WMSYS                          WM$EVENT_QUEUE_TABLE           ALL 
SYSMAN                         MGMT_TASK_QTABLE               ALL 
SYSMAN                         MGMT_NOTIFY_QTABLE             ALL 
SYSMAN                         MGMT_NOTIFY_INPUT_QTABLE       ALL 
SYSMAN                         MGMT_PAF_MSG_QTABLE_1          ALL 
SYSMAN                         MGMT_PAF_MSG_QTABLE_2          ALL 
SYSMAN                         MGMT_LOADER_QTABLE             ALL 
RECOTEST                       BONUS                          ALL 
RECOTEST                       SALGRADE                       ALL 
RECOTEST                       DEPT                           ALL 
RECOTEST                       SPORTS                         ALL 
RECOTEST                       SPORTS_INFORM                  ALL 
RECOTEST                       COPY_T                         ALL 
RECOTEST                       EMP_KIND1                      ALL 
RECOTEST                       EMP_KIND2                      ALL 
SYS                            KUPC$DATAPUMP_QUETAB_1         ALL 
IX                             ORDERS_QUEUETABLE              ALL 
IX                             STREAMS_QUEUE_TABLE            ALL 
SYS                            WRH$_MVPARAMETER               ALL 
SYS                            WRH$_EVENT_HISTOGRAM           ALL 
SYS                            WRH$_SERVICE_WAIT_CLASS        ALL 
SYS                            WRH$_SYS_TIME_MODEL            ALL 
SYS                            WRH$_OSSTAT                    ALL 
SYS                            WRH$_TABLESPACE_STAT           ALL 
SYS                            WRH$_ACTIVE_SESSION_HISTORY    ALL 
SYS                            WRH$_SERVICE_STAT              ALL 
SYS                            WRH$_SEG_STAT                  ALL 
SYS                            WRH$_PARAMETER                 ALL 
SYS                            WRH$_SYSSTAT                   ALL 
SYS                            WRH$_SGASTAT                   ALL 
SYS                            WRH$_ROWCACHE_SUMMARY          ALL 
SYS                            WRH$_DB_CACHE_ADVICE           ALL 
SYS                            WRH$_LATCH_MISSES_SUMMARY      ALL 
SYS                            WRH$_LATCH                     ALL 
SYS                            WRH$_WAITSTAT                  ALL 
SYS                            WRH$_SYSTEM_EVENT              ALL 
SYS                            WRH$_SQLSTAT                   ALL 
SYS                            WRH$_FILESTATXS                ALL 
50 rows selected.

table 별로 해도 되지만 귀찮으므로 schema 한방에 처리하도록 한다

exec DBMS_STATS.UNLOCK_schema_STATS('RECOTEST'); 

-- DBMS_STATS.UNLOCK_[SCHEMA|TABLE]_STATS

해당 schema 에 대한 통계정보를 재수집 하도록 하자

EXEC DBMS_STATS.GATHER_schema_STATS('RECOTEST'); 

-- 잘 안될때에는 force=> true option을 주도록 하자

SQL> show user 

USER is "RECOTEST"

정리되었다- 끝!

SQL> select num_rows from user_tables; 

  NUM_ROWS 
---------- 
         0 
         0 
         0 
         0 
         0 
         0 
         0 
         0 
         0 

9 rows selected.

'6. 먹고살기 > 6.1 Oracle' 카테고리의 다른 글

Hadoop and NoSQL Technologies and the Oracle Database (0)	2012/05/21
object statistics are locked (0)	2012/03/30
COMMIT_POINT_STRENGTH (0)	2012/01/19
11g compress table (0)	2011/09/14

COMMIT_POINT_STRENGTH

주난군 — Thu, 19 Jan 2012 19:25:01 +0900

COMMIT_POINT_STRENGTH

Property	Description
Parameter type	Integer
Default value	`1`
Modifiable	No
Range of values	`0` to `255`
Basic	No

commit_point_strength 분산데이터 시스템에서 사용하는 파라메터이며, 분산 데이터 베이스의 커밋 시점(commit point site)을 결정한다. 해당 값을 가장 높게 설정한 인스턴스가 커밋 시점이 된다(commit point site) 예를 들어 메인프레임과 - 개발자 pc 간의 설정에선 메인프레임의 해당 파라메터값이 높아야함.

'6. 먹고살기 > 6.1 Oracle' 카테고리의 다른 글

object statistics are locked (0)	2012/03/30
COMMIT_POINT_STRENGTH (0)	2012/01/19
11g compress table (0)	2011/09/14
oracle user expired & lock (0)	2011/09/09

잠시..

주난군 — Thu, 01 Dec 2011 21:58:04 +0900

잠시 일을 멈춘다.. 톡톡 거리던 마우스도 잠시 내려놓는다..
저녁밥 먹고 와서 하던 일을 한순간에 멈추었다. 아니 멈추어짐을 당하였다.
원인모를 에러로 파워포인트는 응답없음을 내보내었다.
그 응답없음에 나는 어떤 반응을 보일까. 같이 응답없음으로 응대한다.
시간이 지나도 살아날줄을 모른다. 지나간 시간때 피웠던 담배 생각이 간절하다.
그 이후로 한번도 피우지 않았지만..

문득 예전 사람들이 보고 싶다. 나의 스물세살 스물두살 스물한살.
그때의 그 사람들은 다 어디 간걸까.
나도 그들을 잊고 살았듯이 그들도 나를 잊고 영화를 보거나 친구를 만나거나 손톱을 깎거나 책을 보거나 수다를 떨거나 거리를 거닐거나 밥을 먹거나 애인을 만나거나 버스를 타거나 혹은. 멍 때린다거나.

혹은.

나처럼 일을 하고 있을지도 모를일이다.

잊고 살아도 상관없으니. 다들 부디 건강했으면 좋겠다.
그래서 우연찮게 만났을때. 참 활짝 웃었으면 좋겠다.

부디.

'2.쓰다 > 2.1 일상' 카테고리의 다른 글

나쁜 티스토리.. (0)	2012/04/08
잠시.. (0)	2011/12/01
암튼.. (2)	2011/09/14
어떤 한사람을 좋아하면 좋아할수록 (0)	2011/03/25

11g compress table

주난군 — Wed, 14 Sep 2011 17:14:05 +0900

11g Compress table을 해보기로함..

declare
lv_cmp_ratio number;
lv_comptype_str varchar2(300);
lv_BLKCNT_CMP number;
lv_BLKCNT_UNCMP number;
lv_ROW_CMP number;
lv_ROW_UNCMP number;
begin
dbms_compression.GET_COMPRESSION_RATIO(
SCRATCHTBSNAME=>'EXAMPLE',
OWNNAME=>'SCOTT',
TABNAME=>'IMSI',
PARTNAME =>null,
COMPTYPE =>2, ---2 means OLTP
BLKCNT_CMP=>lv_BLKCNT_CMP,
BLKCNT_UNCMP=>lv_BLKCNT_UNCMP,
ROW_CMP =>lv_ROW_CMP,
ROW_UNCMP =>lv_ROW_UNCMP,
CMP_RATIO=>lv_cmp_ratio,
COMPTYPE_STR=>lv_COMPTYPE_STR);

dbms_output.put_line('====================================================');
dbms_output.put_line('1. Compression Ratio :'||lv_cmp_ratio);
dbms_output.put_line('2. Block Count :'||lv_blkcnt_cmp);
dbms_output.put_line('3. Compression Type :'||lv_comptype_str);
dbms_output.put_line('4. Blk Count Compressed :'||lv_BLKCNT_CMP);
dbms_output.put_line('5. Blk Count Un-compressed:'||lv_BLKCNT_UNCMP);
dbms_output.put_line('6. Row Count Compressed :'||lv_row_cmp);
dbms_output.put_line('4. Row Count Un-Compressed:'||lv_row_uncmp);
dbms_output.put_line('====================================================');
end;
/

적절한 값을 넣은후 실행을 시키면 다음과 같은 값이 나온당

====================================================
1. Compression Ratio :2.8
2. Block Count :694
3. Compression Type :"Compress For OLTP"
4. Blk Count Compressed :694
5. Blk Count Un-compressed:1988
6. Row Count Compressed :202
4. Row Count Un-Compressed:70
====================================================

해당 패키지는 압축률을 계산해 어떠한 방식으로 압축을 했을때

얼마나 효과가 있는지 알려준다 다음은 CTAS로 넘긴후 계산된결과이다

col segment for a40
select segment_name, bytes/1024/1024 as MB, a.blocks, compression, compress_for from dba_segments a,
dba_tables b
where a.segment_name = b.table_name and a.segment_name in ('IMSI','IMSI_COMPRESS');

SEGMENT_NAME                 MB     BLOCKS COMPRESS COMPRESS_FOR
-------------------- ---------- ---------- -------- ------------
IMSI                       4156     531968 DISABLED
IMSI_COMPRESS              1351     172928 ENABLED  BASIC

2 rows selected.

Elapsed: 00:00:00.03

17:08:23 SCOTT@RAC1> select count(*) from imsi;

  COUNT(*)
----------
  37377536

1 row selected.

Elapsed: 00:01:02.38
17:14:11 SCOTT@RAC1> select count(*) from imsi_compress;

  COUNT(*)
----------
  37377536

1 row selected.

Elapsed: 00:00:16.67

예상치보다 압축률이 잘나왔다. 해당 더미데이터는 dba_objects 들이다...

그리고 단순히 count 함수를 써서 돌려보았다..

압축되지 않은 테이블은 1분가량 시간이 소요되었고 압축된 테이블은 16초가량이 걸렸다

09:21:56 SCOTT@RAC1> desc imsi
 Name                                                  Null?    Type
 ----------------------------------------------------- -------- ------------------------------------
 OWNER                                                          VARCHAR2(30)
 OBJECT_NAME                                                    VARCHAR2(128)
 SUBOBJECT_NAME                                                 VARCHAR2(30)
 OBJECT_ID                                                      NUMBER
 DATA_OBJECT_ID                                                 NUMBER
 OBJECT_TYPE                                                    VARCHAR2(19)
 CREATED                                                        DATE
 LAST_DDL_TIME                                                  DATE
 TIMESTAMP                                                      VARCHAR2(19)
 STATUS                                                         VARCHAR2(7)
 TEMPORARY                                                      VARCHAR2(1)
 GENERATED                                                      VARCHAR2(1)
 SECONDARY                                                      VARCHAR2(1)
 NAMESPACE                                                      NUMBER
 EDITION_NAME                                                   VARCHAR2(30)

09:22:34 SCOTT@RAC1> desc imsi_compress
 Name                                                  Null?    Type
 ----------------------------------------------------- -------- ------------------------------------
 OWNER                                                          VARCHAR2(30)
 OBJECT_NAME                                                    VARCHAR2(128)
 SUBOBJECT_NAME                                                 VARCHAR2(30)
 OBJECT_ID                                                      NUMBER
 DATA_OBJECT_ID                                                 NUMBER
 OBJECT_TYPE                                                    VARCHAR2(19)
 CREATED                                                        DATE
 LAST_DDL_TIME                                                  DATE
 TIMESTAMP                                                      VARCHAR2(19)
 STATUS                                                         VARCHAR2(7)
 TEMPORARY                                                      VARCHAR2(1)
 GENERATED                                                      VARCHAR2(1)
 SECONDARY                                                      VARCHAR2(1)
 NAMESPACE                                                      NUMBER
 EDITION_NAME                                                   VARCHAR2(30)

혹시나 해서 memory flush 후 재실행-

17:50:51 SCOTT@RAC1> alter system flush shared_pool;

System altered.

Elapsed: 00:00:23.09
17:51:20 SCOTT@RAC1> alter system flush buffer_cache;

System altered.

Elapsed: 00:00:00.52
17:51:40 SCOTT@RAC1> select count(*) from imsi;

  COUNT(*)
----------
  37377536

1 row selected.

Elapsed: 00:01:10.34

17:53:03 SCOTT@RAC1> alter system flush shared_pool;

System altered.

Elapsed: 00:00:01.99
17:53:53 SCOTT@RAC1> alter system flush buffer_cache;

System altered.

Elapsed: 00:00:00.29
17:53:55 SCOTT@RAC1>  select count(*) from imsi_compress;

  COUNT(*)
----------
  37377536

1 row selected.

Elapsed: 00:00:17.07

값이 줄어든것을 확인할수 있다

'6. 먹고살기 > 6.1 Oracle' 카테고리의 다른 글

COMMIT_POINT_STRENGTH (0)	2012/01/19
11g compress table (0)	2011/09/14
oracle user expired & lock (0)	2011/09/09
Sqlplus, glogin.sql (0)	2011/09/06

암튼..

주난군 — Wed, 14 Sep 2011 08:38:24 +0900

"아버지를 뛰어넘기 위한 아들의 모습" 은 종종 다른 모습으로 비추어 지곤한다. 서양이나 미드에서는 개인적인 컴플렉스의 극복으로 보여지는 반면에 국내 드라마나 우리의 정서상에는 가끔 인간대 인간으로써의 예의 정도로 비추어 지는경우가 많다. 전자와 후자의 차이는 상당히 크다고 할수 있다. 어떠한 한 인간으로서의 완성도를 따진다는 기준에서는 말이다. 덧붙여 부모님의 욕망의 투영이 일어나고 아이도 자연스레 그 타인의 욕망을 욕망하는 수준이 되면, 나중에 어떤 일정한 시점이 되었을때 그 개인적 허무함은 극에 달한다. 뭔가 열씸히 쫓아 왔지만 아무것도 느낄수도 만질수도 없는 상태, 결코 타인의 욕망은 자신의 욕망이 되진 못한다. 저 부분이 미묘하게 교집합을 이루는경우가 종종 있는듯함.

'2.쓰다 > 2.1 일상' 카테고리의 다른 글

잠시.. (0)	2011/12/01
암튼.. (2)	2011/09/14
어떤 한사람을 좋아하면 좋아할수록 (0)	2011/03/25
얼마전 이야기. (2)	2010/10/08

oracle user expired & lock

주난군 — Fri, 09 Sep 2011 14:56:20 +0900

11g 부터 user가 기본으로 180일 제한이 걸려있다..

오라클형들은 자꾸 기본 보안을 높이넹 ~_~

일단 dba_profiles 테이블을 보자면 다음과 같이 되어있음을 확인할수 있다.

10:14:24 SYS@RAC1> desc dba_profiles;
 Name                                                  Null?    Type
 ----------------------------------------------------- -------- ------------------------------------
 PROFILE                                               NOT NULL VARCHAR2(30)
 RESOURCE_NAME                                         NOT NULL VARCHAR2(32)
 RESOURCE_TYPE                                                  VARCHAR2(8)
 LIMIT                                                          VARCHAR2(40)

다음과 같은 명령으로 lock혹은 expired 되어있는 user와 profile을 확인한다

15:04:19 SYS@RAC1> select USERNAME, ACCOUNT_STATUS,PROFILE 15:06:44 2 from dba_users; USERNAME ACCOUNT_STATUS PROFILE ------------------------------ -------------------------------- ------------------------------ BI OPEN DEFAULT SCOTT OPEN DEFAULT OGGT OPEN DEFAULT

Lock 되어있는 유저는 다음과 같은 명령으로 해당 사용자를 unlock 할수 있다.

10:49:40 SYS@RAC1> alter user bi account unlock;
User altered.
Elapsed: 00:00:00.28
11:17:00 SYS@RAC1> select USERNAME, ACCOUNT_STATUS,PROFILE from dba_users where USERNAME='BI';
USERNAME               ACCOUNT_STATUS                   PROFILE
------------------------- -------------------------------- ------------------------------
BI                             EXPIRED                          DEFAULT
1 row selected.

Lock 상태는 사라지고 Expired 상태로 된 것을 확인

13:46:48 SYS@RAC1> select name, password from user$ where name='BI';
NAME                           PASSWORD
------------------------------ ------------------------------
BI                             EB32A21961929D0C
해당 명령어로 user의 password hash값 확인
13:48:17 SYS@RAC1> alter user bi identified by values 'EB32A21961929D0C';
User altered.
13:48:51 SYS@RAC1> select USERNAME, ACCOUNT_STATUS,PROFILE from dba_users where USERNAME='BI';
USERNAME                ACCOUNT_STATUS                   PROFILE
----------------------- -------------------------------- ------------------------------
BI                             OPEN                             DEFAULT
1 row selected.

계정이 Open 상태로 된 것을 확인한다.

이렇게만 하면 일시적으로 expired 되거나 unlock된 계정을 풀수는 있지만...

장기간 운영해하는 입장에서는 여간 귀찮은게 아니다...

default profile을 다음과 같이 수정해서 귀차니즘에서 탈피~!

10:14:12 SYS@RAC1> col limit for a20
select * from dba_profiles where resource_type='PASSWORD';
PROFILE                        RESOURCE_NAME                    RESOURCE LIMIT
------------------------------ -------------------------------- -------- --------------------
MONITORING_PROFILE             FAILED_LOGIN_ATTEMPTS            PASSWORD UNLIMITED
DEFAULT                        FAILED_LOGIN_ATTEMPTS            PASSWORD 10
MONITORING_PROFILE             PASSWORD_LIFE_TIME               PASSWORD DEFAULT
DEFAULT                        PASSWORD_LIFE_TIME               PASSWORD 180
MONITORING_PROFILE             PASSWORD_REUSE_TIME              PASSWORD DEFAULT
DEFAULT                        PASSWORD_REUSE_TIME              PASSWORD UNLIMITED
MONITORING_PROFILE             PASSWORD_REUSE_MAX               PASSWORD DEFAULT
DEFAULT                        PASSWORD_REUSE_MAX               PASSWORD UNLIMITED
MONITORING_PROFILE             PASSWORD_VERIFY_FUNCTION         PASSWORD DEFAULT
DEFAULT                        PASSWORD_VERIFY_FUNCTION         PASSWORD NULL
MONITORING_PROFILE             PASSWORD_LOCK_TIME               PASSWORD DEFAULT
DEFAULT                        PASSWORD_LOCK_TIME               PASSWORD 1
MONITORING_PROFILE             PASSWORD_GRACE_TIME              PASSWORD DEFAULT
DEFAULT                        PASSWORD_GRACE_TIME              PASSWORD 7
SQL>alter profile default limit FAILED_LOGIN_ATTEMPTS unlimited
PASSWORD_LOCK_TIME unlimited
PASSWORD_GRACE_TIME unlimited
PASSWORD_LIFE_TIME unlimited;

패스워드 제한설정

PASSWORD history : 이전 암호를 기억해 놓았다가 다음에 변경시 동일한 암호사용을 금지함 password_reuse_time : 동일한 password를 적용한 기간동안 사용금지

password_reuse_max : 입력된 value값만큼만 사용가능한 횟수를 제한 ex: 3이라고 입력하면 3번만 사용가능 password_life_time : password 생명주기 ex : 30 -> 30일마다 변경해야함

password_grace_time : password 변경 만료 알림을 value일 전부터 알림

failed_login_attempts :password 입력실패시 재시도 가능횟수 최종 실패시 계정 lock걸림

password_lock_time : lock걸렸을때 value값만큼 잠겨있음 1일단위임 1/24는 1시간 1/1440은 1분

password complexity verification(패스워드 복잡성)password설정시 제약조건

'6. 먹고살기 > 6.1 Oracle' 카테고리의 다른 글

COMMIT_POINT_STRENGTH (0)	2012/01/19
11g compress table (0)	2011/09/14
oracle user expired & lock (0)	2011/09/09
Sqlplus, glogin.sql (0)	2011/09/06

Sqlplus, glogin.sql

주난군 — Tue, 06 Sep 2011 14:19:56 +0900

맨날 백업을 안해두어서 생각난김에..

filepath : ?/sqlplus/admin/glogin.sql

set lines 120
set pages 900
define_editor=vi

set verify off termout off head off feed off

col login_prompt new_value welcome

SELECT upper(SYS_CONTEXT('USERENV','SERVER_HOST')
||' '
|| SYS_CONTEXT('USERENV','CURRENT_USER')
||'@'
|| SYS_CONTEXT('USERENV','DB_NAME')) login_prompt
FROM DUAL
;


set sqlprompt "_user'@'_connect_identifier> "

set verify on termout on head on feed on
prompt ************************************
prompt WELCOME TO &&welcome
prompt ************************************
prompt
set echo off serveroutput on size 100000 line 100 trims on
set time on
set timing on

'6. 먹고살기 > 6.1 Oracle' 카테고리의 다른 글

COMMIT_POINT_STRENGTH (0)	2012/01/19
11g compress table (0)	2011/09/14
oracle user expired & lock (0)	2011/09/09
Sqlplus, glogin.sql (0)	2011/09/06

syntaxhighlighter

주난군 — Tue, 06 Sep 2011 13:13:02 +0900

한참까먹고 있다가.. 오늘에서야 적용..

http://gyuha.tistory.com/405 를 참조하였다.

-- active session list

col sid for 999
col sql_text for a20
col event for a10
col sid new_value sid
 select
        to_char(h.sample_time,'mi:ss') as sample_time,
        h.session_id as sid,
        h.session_state as st,
        h.sql_id,
        (select sql_text from v$sqlarea a where a.sql_id = h.sql_id) as sql_text, event,
        blocking_session as blocker
 from
        v$active_session_history h
 where
        sample_id = (select max(sample_id) from v$active_session_history)
        and session_id <> userenv('sid')
;

사용방법은

http://blueray21.tistory.com/30
참조-!

지원 문법

Brush name	Brush aliases	File name
ActionScript3	as3, actionscript3	shBrushAS3.js
Bash/shell	bash, shell	shBrushBash.js
ColdFusion	cf, coldfusion	shBrushColdFusion.js
C#	c-sharp, csharp	shBrushCSharp.js
C++	cpp, c	shBrushCpp.js
CSS	css	shBrushCss.js
Delphi	delphi, pas, pascal	shBrushDelphi.js
Diff	diff, patch	shBrushDiff.js
Erlang	erl, erlang	shBrushErlang.js
Groovy	groovy	shBrushGroovy.js
JavaScript	js, jscript, javascript	shBrushJScript.js
Java	java	shBrushJava.js
JavaFX	jfx, javafx	shBrushJavaFX.js
Perl	perl, pl	shBrushPerl.js
PHP	php	shBrushPhp.js
Plain Text	plain, text	shBrushPlain.js
PowerShell	ps, powershell	shBrushPowerShell.js
Python	py, python	shBrushPython.js
Ruby	rails, ror, ruby	shBrushRuby.js
Scala	scala	shBrushScala.js
SQL	sql	shBrushSql.js
Visual Basic	vb, vbnet	shBrushVb.js
XML	xml, xhtml, xslt, html, xhtml	shBrushXml.js

'6. 먹고살기 > 6.2 기타' 카테고리의 다른 글

syntaxhighlighter (0)	2011/09/06

땡이

주난군 — Fri, 12 Aug 2011 09:01:39 +0900

꼭꼭 숨어랑-
이빨빠져서 힘들어하는 땡이

iPhone 에서 작성된 글입니다.

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

나들잉- (0)	2012/04/22
땡이 (2)	2011/08/12
살아있어서 (0)	2011/06/29
@집 (0)	2011/01/28

살아있어서

주난군 — Wed, 29 Jun 2011 22:28:30 +0900

iPhone 에서 작성된 글입니다.

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

땡이 (2)	2011/08/12
살아있어서 (0)	2011/06/29
@집 (0)	2011/01/28
나는.... (2)	2010/12/25

@콩불_냠

주난군 — Fri, 20 May 2011 12:56:03 +0900

사는게 자신이 없다..

iPhone 에서 작성된 글입니다.

'5.안녕하루' 카테고리의 다른 글

@콩불_냠 (0)	2011/05/20

책____

주난군 — Wed, 20 Apr 2011 22:48:25 +0900

V For 깜박 속았다. 소설책인줄 알았는데!!!!

'1.보다 > 1.1 책. 그리고, 밑줄' 카테고리의 다른 글

책 지름- (0)	2012/04/08
책____ (0)	2011/04/20
@공씨책방 (0)	2011/02/18
책 (2)	2010/11/14

그녀의 손

주난군 — Thu, 07 Apr 2011 19:06:02 +0900

미친듯이 심장이 뛰고 거센 숨이 나를 들었다가 들어내린다. 가끔 이상한 꿈을 꾸곤한다. 끔찍하게 끈적하고 기분나쁜꿈. 막상 그런 꿈을 꾸고 나면 항상 기분나쁜 찝찝함만이 나를 기다리고 있을 뿐이다. 꿈을 꾸었던 자체는 이미 깨끗하게 도려내어지고 완벽한 풍화 작용이 일어나 곧 그런사실이 있었는지 조차 망각하게 된다. 다만 미친듯이 뛰고 있는 심장만이 방금전의 끔찍함을 적나라하게 설명하고 있을뿐이었다. 완전한 내것에서 오로지 나만 느낄수 있는 방법으로. 그럴 때면 나는 항상 그녀의 손을 생각하곤 했다 말갛고 윤기있는 그손을 말이다. 그러면 어느순간에 스스로 차분해져 모든것이 정상으로 돌아오곤 했다. 마치 늘어졌던 용수철이 스스로 되돌아오듯. 어떤 비밀스런 집단의 의식처럼 제의처럼. 항상 그 순간마다 나는 그 손을 생각해온것이다. 손으로 인해 구원 받는것이 아닌 안도감과 평상의 존재확인으로. 그리고 나는 그렇게 그런 이유로 그 손을 사랑하게 되었다.

iPhone 에서 작성된 글입니다.

'2.쓰다 > 2.2 습작' 카테고리의 다른 글

그녀의 손 (0)	2011/04/07
Nobody... (0)	2010/07/21
it's me (2)	2010/07/19
+_+) (5)	2010/06/07

어떤 한사람을 좋아하면 좋아할수록

주난군 — Fri, 25 Mar 2011 00:05:37 +0900

어떤 한사람을 좋아하면 좋아할수록, 어떠한 일에 빠져들면 빠질수록, 사랑이든 일이든. 하나에 너무 몰두하다보면 내가 정말 부족한 사람이라는것을 뼈저리게 느낀다. 노력해서 뛰어넘을수 없는. 어떠한 감각의 부재라던가. "누구나 그러하듯이"의 그런 인간이라면 가장 갖춰야할 기본적인 소양같은것들 말이다. 이를테면 무의식적으로 침을 삼키거나, 눈꺼풀 깜박이는걸 의지로 하지 않듯이.. 경험으로써 으레 그래야 할것들. "사람을 사람으로서 사랑하는것" 이라던가. 흔히 말하는 어떠한 상황에서의 자연스러운 행동들. 그러니까 누가 아프면 약을 사다준다거나. 아니면 최소한의 걱정스러운 눈빛으로 상대방을 봐줄수 있는 그런것들 말이다. 남들에겐 당연한 것들이 나에게 당연하지 않고, 그 당연하지 않은 삶들을 내가 깨달았을때 참 많은것들이 멀게만 느껴진다. 삶은 계속 안좋은 쪽으로만 (혹은 나쁜) 방향으로 강화만 되어가고 말이다. 스무살때 하지못했던것들. 열살때 해보지 못한것들에 대해서. 처음부터 다시 배우자니. 그렇게 힘이 든다. 남들은 당연한것들. 나는 그렇지 않은것들. 어쩌면 숨쉬는법이나, 걸어다니느것에 대해서 난 처음부터 배워야 할지도 모르겠다.

'2.쓰다 > 2.1 일상' 카테고리의 다른 글

암튼.. (2)	2011/09/14
어떤 한사람을 좋아하면 좋아할수록 (0)	2011/03/25
얼마전 이야기. (2)	2010/10/08
이상함! (0)	2010/08/24

@공씨책방

주난군 — Fri, 18 Feb 2011 08:23:10 +0900

블로그에 글을 하도 안썼더니 어색할지경이다....

iPhone 에서 작성된 글입니다.

'1.보다 > 1.1 책. 그리고, 밑줄' 카테고리의 다른 글

책____ (0)	2011/04/20
@공씨책방 (0)	2011/02/18
책 (2)	2010/11/14
팡세 (0)	2010/11/07

@집

주난군 — Fri, 28 Jan 2011 16:26:26 +0900

땡님의 자태-!

iPhone 에서 작성된 글입니다.

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

살아있어서 (0)	2011/06/29
@집 (0)	2011/01/28
나는.... (2)	2010/12/25
@대전 (2)	2010/12/11

나는....

주난군 — Sat, 25 Dec 2010 16:45:11 +0900

iPhone 에서 작성된 글입니다.

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

@집 (0)	2011/01/28
나는.... (2)	2010/12/25
@대전 (2)	2010/12/11
@홍대 (4)	2010/11/20

@대전

주난군 — Sat, 11 Dec 2010 11:53:11 +0900

iPhone 에서 작성된 글입니다.

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

나는.... (2)	2010/12/25
@대전 (2)	2010/12/11
@홍대 (4)	2010/11/20
@노량진 (0)	2010/10/15

@홍대

주난군 — Sat, 20 Nov 2010 23:22:43 +0900

나는 요즘 가끔 웃는다
그리고 많이 운다

예전엔 웃지도 울지도 않았는데.
뭐가 좋은건지는 모르겠다..

iPhone 에서 작성된 글입니다.

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

@대전 (2)	2010/12/11
@홍대 (4)	2010/11/20
@노량진 (0)	2010/10/15
@목포 (2)	2010/10/14

책

주난군 — Sun, 14 Nov 2010 09:54:18 +0900

나의 영원한.
박민규

iPhone 에서 작성된 글입니다.

'1.보다 > 1.1 책. 그리고, 밑줄' 카테고리의 다른 글

@공씨책방 (0)	2011/02/18
책 (2)	2010/11/14
팡세 (0)	2010/11/07
세계의 끝과 하드보일드 원더랜드 - 무라카미 하루키 (0)	2010/09/23

팡세

주난군 — Sun, 07 Nov 2010 17:07:05 +0900

난 요즘 되도 않게 팡세를 읽는다

iPhone 에서 작성된 글입니다.

'1.보다 > 1.1 책. 그리고, 밑줄' 카테고리의 다른 글

책 (2)	2010/11/14
팡세 (0)	2010/11/07
세계의 끝과 하드보일드 원더랜드 - 무라카미 하루키 (0)	2010/09/23
달리기를 말할 때 내가 하고 싶은 이야기 - 무라카미 하루키 (2)	2010/09/23

@노량진

주난군 — Fri, 15 Oct 2010 17:49:40 +0900

정말 오늘 서울 맑음은 HD급!!

iPhone 에서 작성된 글입니다.

'1.보다 > 1.3 파편적기억' 카테고리의 다른 글

@홍대 (4)	2010/11/20
@노량진 (0)	2010/10/15
@목포 (2)	2010/10/14
@광주송정 (0)	2010/10/09