Upload
others
View
25
Download
0
Embed Size (px)
Citation preview
TimesTen OverviewContent & IntentTimesTen OverviewContent & Intent
What is TimesTen?1
Performance2
TimesTen VS RDBMS3
Product Overview4
From backend to edge5
- 2 -
TimesTen OverviewI. What is TimesTen?TimesTen OverviewI. What is TimesTen?
I What is TimesTen ?I. What is TimesTen ?
- 3 -
TimesTen OverviewI. What is TimesTen?TimesTen OverviewI. What is TimesTen?
IMDB란?
Database on Main Memory• 시스템 구동 시 DB 전체를 메인 메모리에 상주
• 메모리 상의 데이터만을 이용하므로 고성능 처리가 가능• 메모리 상의 데이터만을 이용하므로 고성능 처리가 가능
IMDB 등장 배경디스크 I/O에 의한 성능 저하• 디스크 I/O에 의한 성능 저하
• 상용화된 64-Bit Systems의 출현
• 메모리 가격 하락
• 빠른 응답을 요구하는 실시간 시스템 증대
Real-Time 초고속 처리에 적합• I/O성능- Hard Disk:5~10ms, S-RAM:8~35ns, D-RAM: 90~120ns
• Disk 기반 DBMS보다 10배 빠른 성능 제공
- 4 -
TimesTen OverviewI. What is TimesTen?TimesTen OverviewI. What is TimesTen?
TimesTen이란?
MMDB product from Oracle• 전 세계 시장 점유율 1위
• 전 세계 500여개 이상의 기업에서 사용 중• 전 세계 500여개 이상의 기업에서 사용 중
• Product Version 8 (2009.1 출시예정)
실시간 데이터 관리 소프트웨어실시간 데이터 관리 소프트웨어
즉각 반응 고 확장성 고 가용성
고성능이 요구되는 Application내의고성능이 요구되는 Application내의이벤트,트랜잭션,데이터 관리
- 5 -
TimesTen OverviewII. PerformanceTimesTen OverviewII. Performance
II PerformanceII. Performance
- 6 -
TimesTen OverviewII. PerformanceTimesTen OverviewII. Performance
Lightning Fast Response
Average Response TimesTimesTen/DataServer 5.1
IBM eServer pSeries 630, 4-CPU, 1.2 GHz POWER4 AIX 5.2, 64-bit
37 3630
35
40
s
100% Updates
20
25
30
osec
ond 100% Inserts
100% Reads
15
5
10
15
Mic
ro
0Database Operation
- 7 -
TimesTen OverviewII. PerformanceTimesTen OverviewII. Performance
Higher Throughput on Less HardwareTransaction Throughput
350,000
Transaction ThroughputTimesTen/DataServer 5.1
IBM eServer pSeries 670, 12-CPU1.1 GHz POWER4 AIX 5.2, 64-bit
314,076
250,000
300,000
ond
100% Reads70-20-10 Mix100% Inserts100% Deletes100% Updates
208,569
150 000
200,000
,
ons
Per
Seco
pRDBMS 100% Reads
1212--CPU CPU SystemSystem
54 856
107,06295,540
50 000
100,000
150,000
Tran
sact
i SystemSystem
54,856
0
50,000
1 2 4 6 8 10 12
N b f A li ti P
- 8 -
Number of Application Processes
TimesTen OverviewII. PerformanceTimesTen OverviewII. Performance
Higher Throughput on Less HardwareTransaction Throughput
200,000
Transaction ThroughputTimesTen/DataServer 5.1Intel Xeon Pentium 4, 4-CPU
3.0 GHz Pentium 4, RedHat Linux 3.0
186,662
140 000
160,000
180,000
ond
100% Reads
70-20-10 Mix
100% Updates
RDBMS 100% Reads44--CPUCPU
93,318100,000
120,000
140,000
ions
per
sec
o 44--CPUCPUSystemSystem
58,925
31 97140,000
60,000
80,000
Tran
sact
31,971
0
20,000
1 2 4
- 9 -
Number of Application Processes
TimesTen OverviewII. PerformanceTimesTen OverviewII. Performance
Telecom Benchmark - Configurations
Oracle• Three-tier architecture• Benchmark application
Home Location Registry (HLR) application
Oracle + TimesTen• Three-tier architecture• Benchmark application• Benchmark application
running in middle tier• Benchmark application running in middle tier
• TimesTen 7.0 embedded in benchmark application
• Updatable cache• Updatable cache containing benchmark data
• Delete call forwarding• Get access data• Get basic subscriber data
1 Delete
3 Select
transactions
E i t Get basic subscriber data• Get new destination• Insert call forwarding• Update subscriber data• Update subscriber location
3 Select
1 Insert
2 Update
• RAM: 2GB• Processors: Intel Xeon 3GHz hyper-threading * 2• OS: Red Hat Enterprise Linux 4.0
Environment
- 10 -
TimesTen OverviewII. PerformanceTimesTen OverviewII. Performance
Telecom Benchmark - Transaction Response Time
- 11 -
TimesTen OverviewII. PerformanceTimesTen OverviewII. Performance
Telecom Benchmark - Transaction Throughput
- 12 -
TimesTen OverviewIII.TimesTen vs. RDBMSTimesTen OverviewIII.TimesTen vs. RDBMS
III TimesTen vs RDBMSIII.TimesTen vs. RDBMS
- 13 -
TimesTen OverviewIII.TimesTen vs. RDBMSTimesTen OverviewIII.TimesTen vs. RDBMS
TimesTen VS Conventional RDBMS
Oracle-TimesTenIn-Memory Database
Conventional RDBMS
ODBC/N ti API
Applications
Connection 부하 (IPC)
Disk-based 추가 기능에ODBC API
ODBC/Native API
Applications
DBMS 엔진TimesTen RDBMS메모리 복사 부하
데이터 주소 변환 부하
대한 부하
버퍼 관리 부하
• Embeddable: 20-30MB
I/O 부하
• Direct-Linked• Memory-Tuned Data Layout• Disks to Ensure Durability/Recovery
- 14 -
TimesTen OverviewIII.TimesTen vs. RDBMSTimesTen OverviewIII.TimesTen vs. RDBMS
Finding a Row of Data in an RDBMS
- 15 -
TimesTen OverviewIII.TimesTen vs. RDBMSTimesTen OverviewIII.TimesTen vs. RDBMS
Finding a Row of Data in TimesTen
- 16 -
TimesTen OverviewIII.TimesTen vs. RDBMSTimesTen OverviewIII.TimesTen vs. RDBMS
TimesTen VS Oracle
Characteristic Oracle TimesTen
Data Model Relational: SQL Relational: SQL
Target Applications Mission-critical Mission-critical
Optimization Disk-centric Memory-centric
Typical deployment Database tier Application tier
Architecture Client/Server Direct data access
Response Times Milliseconds Microseconds
Data Capacity Tens of terabytes Tens of gigabytes
Scalability Unlimited SMP/cluster Good SMP
- 17 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
IV Product OverviewIV. Product Overview
- 18 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
C h tiR li ti ti
TimesTen 제품군
TimesTen IMDB <Base Product>
Real-Time Database SystemReal-time data management for applications requiring
Cache option
Real-Time Dynamic Data Cache Real-time relational processing of selected data from an Oracle
Replication optionReal-Time Database Systemwith ReplicationReal-time data management, and replication for applications requiring for applications requiring
exceptional performance
Network
selected data from an Oracle database, including automatic data loading and update synchronization
replication, for applications requiring exceptional performance, with the failover, recovery, and evolution of databases with minimal downtime
N t kNetwork
B
Network
CacheDataServer
Typical Uses:Typical Uses:
Primary Subscriber(s)
B
DataServer DataServer
Typical Uses: yp
Application Server Acceleration
Enterprise App Acceleration
Application-Tier Data Caching
yp
Network Equipment & Systems
Value-Added Services
Reference Data
yp
Network Equipment & Systems
Value-Added Services
Reference Data
- 19 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Base Product: TimesTen 구성 요소
A li ti T lApplication or Tool
Web browser thin client
Oracle connect administrator
Application routines
TimesTen Client driver
Serverdaemonor
m
Network
Replication agent(s)
In-MemoryAdmin/Utility
programsApplication
routinesSer
ver P
latfo Server
proxiesagent(s)
Data Store(s)
Ti T
Data Store subdaemon(s)
TimesTenshared libraries
plic
atio
n-Ti
er
Checkpoint files
TimesTen daemonAp
p
Log files
Cacheagent
Oracle D t b
- 20 -
Database
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Base Product: Lifecycle of a TimesTen Datastore1) Data Store image read from Checkpoint file at start up1) Data Store image read from Checkpoint file at start-up2) Transaction written to log file (if configured)3) Changed pages written to oldest checkpoint file4) For safety, checkpoints are written to two alternating files5) A final Checkpoint is written at shut-down
DataStore(1) Upon Start-up and/or
Recovery Upon Exit
(5)
TimesTen
Recovery
(4)
Upon Exit
TimesTen
FullData Checkpoint
FullData Checkpoint
Upon Commit O f
(2)
Upon Checkpoint “n+1”(4)
Upon Checkpoint “n”(3)
All Changed PagesSince last “B” Checkpoint
StoreImage
CheckpointFile “B”
All Changed PagesSince last “A” Checkpoint
StoreImage
CheckpointFile “A”
…Inserts,Updates,Deletes
Log
Or flush
- 21 -
Since last checkpoint
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Base Product: Features
RDBMS/SQL 데이터 모델
• JDBC, ODBC APIs
• API for change notification Application
Network
TimesTen
Application
API for change notification
Application tier execution실시간 성능
TimesTen Libraries
Application
TimesTen Libraries
Application
Client-Server
TimesTen Libraries
ApplicationTimesTen Client lib
• Low latency and high throughput
• Efficient CPU utilization
데이터 지속성 및 복구 (on disk)
ServerDirect-linked
In-Memory데이터 지속성 및 복구 (on disk)유연한 구성
• Async and sync disk loggingTransaction Logs
In-Memory Database
• Embedded and client-server
관리 영역이 거의 없음Checkpoint files
Transaction Logs
- 22 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Base Product: 동시성 제어
Data Store 접근은 다중 프로세스와 Thread에 의해 공유됨
• Direct Data Access mode (low latency, very high-performance, database tightly coupled to application processes, embedded)
• Client-server mode (less performance intensive applications)
Concurrency control • Locks: data-store-level, table-level, row-level• Latches: 내부 데이터 구조에 대한 locking
Least
Most
Transaction isolation• Read committed • Serializable
ola
tion
ncurr
encySerializabl
e
t
자동적으로 Deadlock 감지 및 처리
Iso
Con
Read Committed Leas Most
- 23 -
t
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Base Product: Transaction ACID
Atomicity : full support for transactions(Commit/Rollback)Consistency
• 데이터의 일관성 유지. 자동적으로 디스크 정보를 이용하여 복구
작업 수행
Isolation (By locking)• Row level locking• Row-level locking
• READ COMMITTED isolation(default)
• 변경이 가해지는 Row들은 versioning 됨
• SERIALIZABLE isolation 또한 가능
Durability• 트랜잭션은 디스크에 기록되고 디스크를 이용해 복구됨• 트랜잭션은 디스크에 기록되고, 디스크를 이용해 복구됨
트랜잭션이 데이터를 변경할 때, logging과 locking이 이용되어 ACID 규칙을 유지하게 됨
- 24 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Replication: TimesTen to TimesTen
Network
Application Application
N-way, peer-to-peer Active/standby or
ti / ti lti l l
n-Ti
er P
latfo
rms Application
Program
TimesTen
Application Program
TimesTen
active/active, multi-levelData Store / 테이블 수준 정의
동기/비동기 전송In-MemoryDatabase
In-MemoryDatabase
App
licat
ion
자동복구
데이터 충돌 감지 및 처리
ReplicationTimesTen to TimesTen
- 25 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Replication: 다양한 구성
Active - Standby N – Way (update anywhere)
Active - Active (split workload)
Propagation
Active - Active (distributed workload)
- 26 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Replication: Async Rep. Inside
(1) 트랜잭션이 커밋, 어플리케이션은Sender Replication Agent 보내기 위한트랜잭션을 모은다.
(2) Full일 때 Sender Agent는 로그 파일에
Sender
(2) Full일 때 Sender Agent는 로그 파일에트랜잭션을 기록하고(durably commit),
(3) Receiver에 전달한다.
Receiver(4) Receiver Agent는 모든 트랜잭션을
data store에 적용하고에 적용하(5) Log file에 기록한다.(6) Receiver Agent 작업 결과를
Primary Agent에 보낸다.
(7) Sender Agent 트랜잭션 로그 파일에bookmark를 한다.
Sender
- 27 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Replication: Sync Rep.(Return Receipt) Inside
Sender
(1) 응답 회수 방식으로 트랜잭션 커밋
(2) Sender Agent는 로그 파일에 트랜잭션을기록하고(durably commits).
에 내용을 전달
Receiver
(3) Receiver에 내용을 전달.
(4) Receiver 수신 신호를 송신.
Sender
(5) 어플리케이션으로 커밋 리턴.
R i(6) Receiver Agent는 모든 트랜잭션을 data
store에 적용하고(7) Log file에 기록.
Receiver
(8) Receiver는 최종 응답을 송신.
Sender(9) Sender Agent 트랜잭션 로그 파일에 bookmark를 기록.
- 28 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Replication: Sync Rep.(Return 2-Safe ) Inside
Sender
(1) 2-Safe방식으로 트랜잭션 커밋
Local txn는 pre-commit 로그 레코드를 기록.
(2) Receiver에 로그를 전달
Receiver
(3) Receiver Agent는 data store에서
트랜잭션 로그를 적용하고,
Receiver
(4) 응답을 보내고
Sender
(5) Bookmark를 수정.
(6) Sender Agent는 대기 중인 트랜잭션을 풀고
(7) Bookmark를 수정.
- 29 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Cache Connect to Oracle
오라클 테이블 전체 또는 부분 데이터를
Caching
Read-only or updateable
양방향 데이터 전송
• From TimesTen to Oracle
• From Oracle to TimesTen
Async or sync write-through
하나의 오라클 DB에 다중 Data Store 가능
Replication option과 연동하여 사용 가능
- 30 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Cache Connect to Oracle
오라클 DB의 앞 단에 고성능 Cache를 위치시키는 효과
오라클과 TimesTen간의 양방향 데이터 전송
10g RAC와 연동
기본 구성요소:
• TimesTen IMDB
• Oracle Agent
C h G• Cache Groups
• 웹 환경 GUI툴
- 31 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Cache Connect to Oracle: 양 시스템간 데이터 전송
양 DB간의 데이터 전송은 Application 로직에 독립적
TimesTen updates to Oracle
• Automatic Propagation or manual Flush
Oracle updates to TimesTen
• Full refresh
• Incremental refresh
A t ti M l
Propagate Refresh
• Automatic or Manual
Cache Group 기준
- 32 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Cache Connect to Oracle: Cache Group
Use Cache Groups to
• 테이블 하나 또는 다중 테이블을 대상으로 정의
• 컬럼 범위 설정 / Row 범위 설정
• 동기화 행위 지정
ReadOnly
Synchronous WriteThrough
A h W it Th hAsynchronous WriteThrough
User Managed
- 33 -
TimesTen OverviewIV.Product OverviewTimesTen OverviewIV.Product Overview
Cache Connect to Oracle: Failover for Oracle RAC
RAC 노드 failure 자동 감지
가용한 노드로 자동 재 접속
- 34 -
TimesTen OverviewV.Support ToolsTimesTen OverviewV.Support Tools
V Support ToolsV. Support Tools
- 35 -
TimesTen OverviewV.Support ToolsTimesTen OverviewV.Support Tools
TimesTen Extension for Oracle SQL Developer
Graphical Development ORACLE Tool
Browse, create, edit, and delete TimesTen database objects
- 36 -
TimesTen OverviewV.Support ToolsTimesTen OverviewV.Support Tools
TimesTen Extension for Oracle SQL Developer
Run SQL statements and scripts
Execute TimesTen built-in procedures and utilities
- 37 -
TimesTen OverviewV.Support ToolsTimesTen OverviewV.Support Tools
TimesTen Plug-In for Oracle Enterprise Manager
Monitor system information
TimesTen configuration and performance statistics
- 38 -
TimesTen OverviewVI.From backend to edgeTimesTen OverviewVI.From backend to edge
VI From Backend to edgeVI. From Backend to edge
- 39 -
TimesTen OverviewVI.From backend to edgeTimesTen OverviewVI.From backend to edge
Where deployed ?
Mobile DevicesTelematicsField Force Automation
Telecom OSS/BSSIP Multimedia Systems
Gateways / RoutersStorage & Systems Mgmt
Healthcare EquipmentDistributed
Asset Management
IP Multimedia SystemsCore Networking
SystemsSecurities Trading
Security / AuthenticationMobile Services Software
Appliances
Healthcare EquipmentIndustrial Systems
Manufacturing SystemsSemiconductor
Call Centers / CRMReservations
EquipmentCommunications
SystemsTelco SoftSwitches OCMS /
ONCS
Mobile Applications
ONCS
Data CenterApplications
Edge Applications
- 40 -