1 copyright © 2012, oracle and/or its affiliates. all ... · better scalability via multi...


Integrate Your Globally Distributed Databases for Key Cloud Computing Benefits Oracle Database 12c Global Data Services Srinagesh Battula Senior Principal Product Manager - Oracle Cris Pedregal Martin CMTS - Oracle Sarah Brydon Database Engineer - PayPal


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.


Challenges of Databases in Distributed Environments

Introduction to Global Data Services (GDS)

GDS Use Cases, Concepts, and Architecture

Attributes of Global Services

GDS Capabilities: Load Balancing and Failover in a Global Scope

Customer’s GDS Case Study - PayPal

Summary

Agenda


Local Standby

Data Center #2 Data Center #1

Active Data Guard

Distributed Environments

RAC Primary Single Instance

Reader Farm

Active Data Guard

Local Standby

Active Data Guard

Local Standby

Active Data Guard

Master Master

Oracle GoldenGate

Active Data Guard

Challenges – No seamless way to

efficiently use all the databases

– No automated load balancing and fault tolerance

– No complete centralized management

Resulting in – Sub-optimal resource

utilization

– Hampered or no enterprise-wide data integration

– Unclear strategy for consolidation versus distribution

Solution A framework that transparently manages client workloads across replicated databases


Global Data Services (GDS)

Based on two foundations: – Services: proven concept for dynamic workload management for RAC – Replication: Active Data Guard, Oracle GoldenGate

What it does: – Extends RAC-style failover, load balancing (within and across data centers),

and management capabilities to database services on replicated databases – Takes into account network latency, replication lag, and service placement

This results in: – Higher Availability via service failover across local / global databases – Better Scalability via multi-database integrated load balancing – Better Manageability via centralized administration of global resources

New availability and scalability feature in Oracle Database 12c


Application Workload Suitability for GDS GDS is best for applications that: Are replication-aware

– Designed to work in replicated environments Can separate their work into read-only and read-write services

– Example: read-only loads on an Active Data Guard standby Are aware of and avoid or resolve update conflicts

– Example: Oracle GoldenGate multi-master configurations Can tolerate some replication lag

– Many reporting applications can work with slightly out-of-date data


Global Data Services Concept

Without GDS

Sales Service Sales Service

GoldenGate GoldenGate

Sales Global Service

With GDS

Unified Framework


GDS Use Cases

With Active Data Guard – Service failover and load balancing within local data center – Service failover and load balancing across data centers – Automatic role-based Services upon Data Guard role transitions – Load balancing for reader farms

With Oracle GoldenGate – Load balancing across data centers

Some Examples


Active Data Guard

Active Data Guard

Physical standby

GDS Framework

Read Only Service

GDS – Use Case with Active Data Guard Service Failover and Load Balancing within Local Data Center

• Read Write Service runs on Primary

• Read Only Services load balanced across all standby databases

• If a Physical Standby fails, GDS fails over Read Only Service to an available database

• Integrate Active Data Guard replicated databases into a scalable and highly available private data cloud

Physical standby

Read Only Service

Load Load

Primary

Read Write Service

Read Only Service

Load

Connection Pool


Active Data Guard

Physical standby

Connection Pool

Active Data Guard

Primary Physical standby

GDS Framework

Connection Pool

Data Center 1 Data Center 2

Read Write Service

Read Only Service

Read Only Service

Read Only Service

GDS – Use Case with Active Data Guard Service Failover and Load Balancing across Data Centers


• Read Only Services load balanced across all standby databases

• If a physical Standby fails, GDS fails over Read Only Service to Primary

• Integrate Active Data Guard replicated databases (local and remote) into a scalable and highly available private data cloud

Load Load Load


Oracle GoldenGate

Connection Pool

GDS Framework

Connection Pool

Data Center 1 Data Center 2

Master

Read Write Service

Master

GDS – Use Case with GoldenGate (Multi-Master) • Application handles

multi-master conflict resolutions

• Read Write Service load balanced on both Masters

• Integrate Oracle GoldenGate replicated databases (local and remote) into a scalable and highly available private data cloud

Load

Load Balancing across Data Centers

Read Write Service

Read Write Service

Read Write Service

Load

Load

Load


Global Data Services - Concepts GDS Region: Group of databases and clients in close network proximity, e.g., East, West GDS Pool: Databases that offer a common set of global services, e.g., HR, Sales Global Service: Database Service provided by multiple databases with replicated data

– Local service + {region affinity, replication lag, database cardinality}

Global Service Manager (GSM): Provides main GDS functionality: service management and load balancing

– Regional listener to the incoming database connections – At least one GSM per region or multiple GSMs for High Availability – All databases/services register to all GSM Listeners

GDS Catalog: stores all metadata, enables centralized global monitoring & management – Global service configuration stored in GDS Catalog

GDSCTL: Command-line Interface to administer GDS


Local Standby

Local Standby

Data Center #2 EMEA

Data Center #1 APAC

Active Data Guard

Active Data Guard

Global Data Services Globally Distributed, High Availability Architecture

RAC Primary

Local Standby

Active Data Guard

Single Instance Reader Farm

Active Data Guard

GDSCTL GSM GSM GSM GSM

GDS Catalog Database

Master

• GDS Framework dynamically balances user requests across multiple replicated sites – Based on location, load,

and availability

• Provides global availability – Supports automatic

service failover

• GDS integrates disparate databases into a unified data cloud

GDS Catalog Standby

Master

Oracle GoldenGate

Active Data Guard

SALES POOL (sales_reporting_srvc, sales_entry_srvc)

HR POOL(hr_apac_srvc, hr_emea_srvc)

All GDS databases connected to all GSMs

GSM - Global Service Manager


GDS Configuration Flow

Cloud Catalog DB

Pool Database

1. GDSCTL connects to catalog DB via GSM listener

2. GSM routes request to catalog

3. GDSCTL establishes connection to catalog DB

4. GDSCTL modifies the catalog

5. Catalog notifies GSMs of changes

Pool Database

6. Master GSM modifies Pool Databases

GSM GSM


Client Connectivity in GDS

(DESCRIPTION=

(FAILOVER=on)

(ADDRESS_LIST=

(LOAD_BALANCE=ON)

(ADDRESS=(global_protocol_address_information))

(ADDRESS=(global_protocol_address_information)))

(ADDRESS_LIST=

(LOAD_BALANCE=ON)

(ADDRESS=(global_protocol_address_information))

(ADDRESS=(global_protocol_address_information)))

(CONNECT_DATA=

(SERVICE_NAME=global_service_name)

(REGION=region_name)))

Clients connect to GSM listener instead of to the database listener GDS forwards the connection to the local

listener (bypassing the SCAN listeners)

GSM listeners endpoints must be specified Client will load balance among local GSMs

and use the remote GSMs if all the local GSMs are unavailable

Clients specify global service name and which region they want to connect

All current FAN events (instance, service, net, node - up/down) are supported

APAC’s GSMs

EMEA’s GSMs


Global Service Manager (GSM) Maintains GDS configuration Manages cardinality and failover of global services Measures network latency between its own and other regions and

exchanges this information with other regions’ GSMs Performs connection load balancing based on Connection Load

Balancing metrics (from database instances), network latency, region Monitors database instances, generates and publishes normalized

Runtime Load Balancing events for clients in local region








Summary

Agenda


Service Placement

Preferred Databases: designated to provide the Global Service – The number of preferred databases defines the cardinality of the Global Service

Available Databases: provide Global Service if not enough Preferred are running

– If one of the preferred databases fails, then GSM maintains the cardinality of the Global service by starting the service on an available database

Preferred_All: All the databases in a Pool are used for the Global Service Role-based Global Service

– Can be started on a database if its role is the global service’s role attribute – Oracle Clusterware not required for this scenario



Replication Lag for Active Data Guard

An asynchronous standby may lag behind its primary database Applications may choose between real-time vs. slightly out-of-date data Applications can set maximum acceptable lag limit for a Global Service GDS routes requests to replicas whose replication lag is below the limit When replication lag exceeds the lag limit, service is brought down New requests are routed to a database that satisfies the lag limit



Region Affinity

For Local-Region affinity, services are restricted to the local region – Regardless of load, GDS will not route to databases in other regions

For Any-Region affinity, client connections and work requests are routed to any region for load balancing or failover For Local with Inter-Regional Failover affinity, client connections and

work requests are routed to another region in the following cases: – Singleton services (for example, only one master site for update) – All databases in a region have failed









Summary

Agenda


Global Service Failover

When a global service fails, GDS will restart it on an available database in the cloud (across regions) Notifications will be sent so that client connection pools can reconnect to new

database – Allows synchronized failover of mid-tiers and applications – Can support all connection pools or applications that can process FAN events

GDS Catalog ensures consistent view of the configuration across the cloud GDS supports role-based global services

– Automatically starts a global service only when db role matches the role specified for the service


Global Service Failover

Sales-2

Connection Pools have established connections to Sales-1 Database with the service SALES

Sales-1 Workloads are redirected to Sales-2 database

Global Service : SALES

App Server

Global Service : SALES

Clients connected to the Application Server

GSM via Oracle Notification Services (ONS) notifies connection pool that service SALES is UP with database Sales-2

GSM finds another database to service SALES

GSM starts service SALES

GSM


Global Connection Load Balancing (CLB)

CLB balances connection requests across GSM listeners – includes connect-time failover

Client tries to connect to any of the local-region GSMs first – If local GSMs don’t respond, client tries a GSM in another region

TNS-entries must contain two lists of addresses: – one list of local GSMs for load balancing and intra-region failover, and – another list of addresses for remote GSMs for inter-region failover

Client-Side Load Balancing


Global Connection Load Balancing (GCLB)

Every GSM receives load statistics from all databases in the GDS pool and measures network latency between regions to provide CLB functionality GSM Listener directs connection requests

to the best database in the GDS cloud Server-side CLB goals

– “LONG” for applications with long-lived connections e.g. connection pools and SQL*Forms sessions

– “SHORT” for applications with short-lived connections

Server-side Load Balancing

DB5 (RAC)

DB1 GSM

Client Connection

Pool

CLB


Client Connections Server-side Load Balancing

Sales-2

1.Client connects to a random GSM in the same region (sales_read_service.salespool.oradbcloud)

2. Based on CLB information, GSM redirects client to an optimal instance.

SHORT, LONG REGION Affinity Instance Performance

Sales-1

All databases connected to all GSMs

Global Service : SALES_READ_SERVICE Global Service : SALES_READ_SERVICE

GSM GSM


Global Runtime Load Balancing (RLB) GDS supports balancing of work requests at runtime for a global

service GSM receives per-service performance data from all database

instances in the cloud – Measures inter-region network latency – Creates an array that specifies what percentage of client requests

each instance should get – Sends the array to regional clients as RLB event via ONS Channels*

Allows client connection pools to route database requests based on real-time load information

* OCI and ODP.NET will be using ONS instead of AQ in 12.1 – simplifying configuration


Client Redirection Run-time Load Balancing

Sales-2

Client has established connection to the service (CLB enabled by GSM) and workloads are directed to this database instance

As a Master GSM, compute and publish FAN RLB events as an array of normalized % of clients to services.

Sales-1

All databases connected to all GSMs and provides performance information

RLB based on SERVICE_TIME, THROUGHPUT, Instance Performance, other GSM parameters

Workload redirected based on RLB

GSM

GSM


GDS Runtime Load Balancing Graph (I) Standalone Identical Database Servers - Simple External Load

Routing responds gracefully to changing system conditions


GDS Runtime Load Balancing Graph (II) Standalone Asymmetrical Database Servers - No External Load DB b: 4 CPUs DB c: 3 CPUs DB d: 2 CPUs

GDS does intelligent load balancing even across asymmetrical database servers








Summary

Agenda

Global Database Services in Oracle Database 12c Sarah Brydon, PayPal Database Engineer

Who am I?

• Oracle/Unix DBA since 1996

• Worked with every Oracle version from 7.1 on

• Oracle Certified Master (and more) – Oracle Certified Professional 7, 8, 8i, 9i, 11g; Managing

Oracle on Linux; Oracle Certified Master

• Specialist in RAC deployments, 24x7 environments, Oracle security

• Member of the Paypal Database Engineering team

34

Tier-1 Oracle Database HA at PayPal (2012)

35

Data Guard Redo Transport

GoldenGate Replication

Active Data Guard Standby • Offload queries – eg balance, other

payment-related queries, industry-specific analytics, etc.

Mission-critical Payment Processing Databases Supporting up to 300K SQL executions per sec

Primary Data Center

Data Guard Cascaded Redo Transport

DR Data Center Data Guard ASYNC Redo Transport

ETL Targets

Production Databases • RAC, ASM, FRA • 10-40 TB

Active Data Guard Standby • Offload queries and reads

Active Dataguard Standby • Supports DR and batch, ETL

WAN, 650+ miles

GoldenGate

ETL Targets

PayPal’s Business Challenge • Support read services on multiple Active Data Guard

databases

• Meet defined SLAs for lag on read-only services

• Manage services in multiple data centers – Balance load across Active Data Guard copies – direct connections to local region

• Service location transparency to clients – Manage service availability during planned maintenance – Relocate primary database and perform tech refreshes ‘in the

cloud’ 36

The Case for Global Data Services

• Simple, centralized management of services – Define the service once in GDS and specify all preferred

and available databases – Data Guard Broker integration for role-aware service

definitions

• Performance management – Specify a maximum lag and the service will automatically

be disabled if the lag is exceeded – Connection and Runtime Load Balancing options – Region affinity for global services with inter-region failover

37

PayPal Lab setup – Oracle Database 12c Beta

38

Data Guard Redo Transport

ASYNC

GDSCAT

ADG STANDBY (LABLNXC) PRIMARY (LABLNXA) ADG STANDBY (LABLNXB)

GSM

11.2 jdbc clients,no UCP 12.1 jdbc thin clients , UCP read connections

11.2 jdbc clients,no UCP 12.1 jdbc thin clients , UCP read connections

Databases ------------------------ lablnxa lablnxb lablnxc Services ------------------------ srv_lablag15 srv_labroregion srv_labrw (primary read-write service)

Region C Region B

Simple, centralized management -- add a service once in GDSCTL and it deploys to every appropriate instance

add service -service srv_lablag15 -gdspool lab -preferred lablnxb,lablnxc -available lablnxa -role PHYSICAL_STANDBY -lag 15 -loadbalance LONG

add service -service srv_roregion -gdspool lab -preferred lablnxb,lablnxc -available lablnxa -role PHYSICAL_STANDBY -loadbalance LONG -locality LOCAL_ONLY -region_failover

GDSCTL>services … Service "srv_lablag15.lab.oradbcloud" has 2 instance(s). Affinity: ANYWHERE Instance "lab%17", name: "LABLNXC", db: "lablnxc", region: "scfc", status: ready. Instance "lab%9", name: "LABLNXB", db: "lablnxb", region: "scfb", status: ready. Service "srv_roregion.lab.oradbcloud" has 2 instance(s). Affinity: LOCALPREF Instance "lab%17", name: "LABLNXC", db: "lablnxc", region: "scfc", status: ready. Instance "lab%9", name: "LABLNXB", db: "lablnxb", region: "scfb", status: ready.

-- location neutral connection strings

jdbc:oracle:thin:@<gsm address and failover settings> (PORT=1571)))(CONNECT_DATA=(SERVICE_NAME=srv_roregion.lab.oradbcloud)(REGION=scfb)))

39

Summary

• True enterprise-wide management of services

• Manage services in the cloud by abstracting database connection strings

• Region-aware services supports the growing need for management of services across databases that may be physically widely separated

• Smart integration with broker configurations to leverage Active Data Guard databases

40








Summary

Agenda


GDS Additional Use Cases

With Active Data Guard – Service failover and load balancing within local data center – Service failover and load balancing across data centers – Automatic role-based services upon Data Guard role transitions – Load balancing for reader farms

With Oracle GoldenGate – Load balancing across data centers

Some Examples


Active Data Guard

Primary Physical standby

GDS Framework

Connection Pool

GDS – Use Case with Active Data Guard Data Guard Role Transition

Read Write Service


• Read Only Service runs on Standby

Read Only Service

- Global Service Management (GSM) Listeners


Active Data Guard

Primary

GDS Framework

Connection Pool

GDS – Use Case with Active Data Guard Data Guard Role Transition

Read Write Service

Read Only Service


• Read Only Service runs on Standby

• Upon Data Guard role change via Broker, GDS fails over Read Write Service to new Primary and Read Only Service to new Standby

• GDS performs automatic role based service management for Data Guard configurations (Oracle Clusterware not needed for this scenario)

Physical standby


GDS – Use Case with Reader farms Load Balancing in an Active Data Guard reader farm

Active Data Guard Reader Farm

Primary

Physical standby

Physical standby

Physical standby

Physical standby

Physical standby

GDS Framework

Connection Pool


• Read Only Services load balanced over standby database reader farm

• Results in effective utilization of Active Data Guard standby databases Read Only

Service Read Only

Service Read Only

Service Read Only

Service Read Only Service

Read Write Service

Load Load Load Load Load


Key Takeaways – Global Data Services

GDS extends RAC-style load balancing, failover, and managing capabilities for distributed environments of replicated databases: – Better Scalability

GDS balances workloads over any set of replicated databases – Higher Availability

GDS can failover Services across local and remote databases – Global Manageability

GDS centralizes management of global resources

GDS is available with Oracle Database 12c Release 1

Copyright © 2012, Oracle and/or its affiliates. All rights reserved. 48 After OpenWorld, visit oracle.com/goto/availability

Key HA Sessions and Demos by Oracle Development Monday, 1 October – Moscone South

12:30p Oracle Data Guard Zero-Data-Loss Protection at Any Distance, 300 12:30p Future of Exadata: OLTP, Warehousing, and Consolidation, 104 1:45p Automating ILM with the Latest Database Technology, 300 1:45p Extracting Data in Oracle GoldenGate Integrated Capture Mode, 102 3:15p Maximize Availability with the Latest Database Technology, 303 3:15p Maximize Enterprise Availability with the Latest DB Technology, 303 4:45p Mission-Critical Oracle Exadata OLTP Deployment at PayPal, 300 4:45p Temporal Database Capabilities with the Latest DB Technology, 300 Tuesday, 2 October – Moscone South 10:15a Database Tables to Storage Bits: Data Protection Best Practices, 300 10:15a GoldenGate & Data Guard: Working Together Seamlessly, 305 11:45a Active Data Guard Zero-Downtime Database Maintenance, 300 11:45a Using Automatic Storage Mgmt with the Latest DB Technology, 301 1:15p The Four Ts of RMAN: Tips, Tuning, Troubleshooting, and … ?, 102 5:00p Maximum Availability Architecture Best Practices for Exadata, 303

Wednesday, 3 October – Moscone South 10:15a Operational Best Practices for Oracle Exadata, 102 10:15a Maximize Availability by Minimizing Disruption for End Users and Application, 301 11:45a What’s New in the Latest Generation of Oracle RAC, 301 11:45a Best Practices for HA w/ GoldenGate on Oracle Exadata, 102 1:15p Oracle Secure Backup: Integration Best Practices with Engineered Systems, 300 1:15p Application MAA Best Practices on Oracle Private Clouds, 200 5:00p Tuning &Troubleshooting Oracle GoldenGate on Oracle, 102 Thursday, 4 October – Moscone South 11:15a Integrate Your Globally Distributed Databases for Key Cloud Computing Benefits, 300 12:45p Backup and Recovery of Oracle Exadata: Experiences and Best Practices, 300

Demos – Mon 10:00a-6:00p - Tue 9:45a-6:00p - Wed 9:45a-4:00p Oracle Maximum Availability Architecture, S-011 GoldenGate 11gR2: Real-Time, Transactional DB Replication, S-027 Oracle Database 12c: Global Data Services, S-010 Oracle Database 12c Application Continuity - S-009

Oracle Secure Backup, S-014 Oracle Active Data Guard, S-007 Oracle Recovery Manager and Oracle Flashback Technologies, S-019 Oracle Real Application Clusters and Oracle RAC One Node - S-008 Oracle Database 12c Xstream, Streams, Advanced Queing, S-018


Resources OTN HA Portal:

http://www.oracle.com/goto/availability

Maximum Availability Architecture (MAA): http://www.oracle.com/goto/maa

MAA Blogs: http://blogs.oracle.com/maa

Exadata on OTN: http://www.oracle.com/technetwork/database/exadata/index.html

Oracle HA Customer Success Stories on OTN: http://www.oracle.com/technetwork/database/features/ha-casestudies-098033.html

1 copyright © 2012, oracle and/or its affiliates. all ... · better scalability via multi...

Documents