Your SlideShare is downloading. ×
Creating a Maximum Availability Architecture with SPARC SuperCluster
Upcoming SlideShare
Loading in...5
×

Thanks for flagging this SlideShare!

Oops! An error has occurred.

×

Saving this for later?

Get the SlideShare app to save on your phone or tablet. Read anywhere, anytime - even offline.

Text the download link to your phone

Standard text messaging rates apply

Creating a Maximum Availability Architecture with SPARC SuperCluster

1,316
views

Published on

Learn how to deploy the integrated capabilities of Oracle’s SPARC SuperCluster to meet the most-demanding requirements for very high levels of availability, scalability, performance, and disaster …

Learn how to deploy the integrated capabilities of Oracle’s SPARC SuperCluster to meet the most-demanding requirements for very high levels of availability, scalability, performance, and disaster recovery. This session discusses how to create a maximum availability architecture with Oracle Real Application Clusters (Oracle RAC), Oracle Active Data Guard, Oracle Database’s Oracle Recovery Manager (Oracle RMAN) feature, Oracle Enterprise Manager Cloud Control, and Oracle Enterprise Manager Ops Center. The session also examines how to leverage the built-in redundancy and security of the SPARC SuperCluster and the ease of migration provided by Oracle Solaris 11 to reduce risk and time to implementation.

Published in: Technology

0 Comments
0 Likes
Statistics
Notes
  • Be the first to comment

  • Be the first to like this

No Downloads
Views
Total Views
1,316
On Slideshare
0
From Embeds
0
Number of Embeds
0
Actions
Shares
0
Downloads
65
Comments
0
Likes
0
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
No notes for slide
  • Data Guard BrokerEasier to manage and control the DG environmentSwitchover, Failover, and Re-Synching procedures are fully documented to allow DBA to execute procedures while under pressure
  • Building all the elaborate backup, snapshot, and Disaster Recovery mechanisms is great, but for true Maximum Availability, we need to start with a solid highly available hardware base.The SPARC SuperCluster starts with all critical external connections being duplicated. Beginning with power, it has dual power distribution units, and each switch, T4-4, DB storage cell, and 7320 ZFS Storage Appliance has connections to each PDU, and of course each PDU must be be connected to a separate power source in the lab.Next, the 10 Gigabit Ethernet networking. Each T4-4 domain in a SuperCluster has two dedicated physical network ports we configure in active/standby IPMP mode for one logical connection. Again each pair is connected to two separate lab network switches outside the SuperCluster.The internal high speed InfiniBand Backbone of the SuperCluster, connecting 11gR2 instances to the Database storage cells, and all T4-4 domains to the 7320 ZFS Storage Appliance, consists of 2 Infiniband Leaf nodes with connections to each of the aforementioned, and an Infiniband spine switch connecting the leaves. This provides not only the required redundancy but also allows for connecting additional Engineered systems or, in this case, a 7420 ZFS backup appliance.Of course we use mirrored local disks for the Solaris 10 and 11 OS instances on the T4-4s, and the T4-4s themselves have many built in HA features, like hot swap PCI-Express cards, memory DIMM off lining.The 7320 ZFS Storage Appliance is composed of two identical Solaris-based head nodes, each with redundant IB network connections and redundant connections to the disk storage tray. And of course one can configure the RAID type and spare options for shares the appliance provides.The Exadata Database Storage Cells also each have redundant InfiniBand network connections and use double or triple disk mirroring.About the only non-redundant hardware in the whole rack is the 1 gigabit management switch and network, though even that could be considered redundant in a manner of speaking since there is a 1 gigabit OS connection and a 1 gigabit service processor connection to each major component in the rack. In any case, even if those both were to fail or the switch were to fail, the major components can all be reached via 10 gigabit network.
  • On top of the High Availability by hardware topology, every SPARC SuperCluster is designed with key Oracle High Availability features. For the Database tier, multiple RAC instances provide continual DB uptime even in the event of DB software, OS, or complete hardware failure on a T4-4. Or even on multiple T4-4s depending upon the SuperCluster configuration chosen.RAC is built on top of Oracle Grid Infrastructure, which abstracts the Database Storage Cells and their disks. This provides continued Database uptime even in the event of disk or even DB Storage node failure.On the application tier, customers may choose their own horizontal scaling and failover mechanisms as was the case for this implementation, or may choose to use Oracle Solaris Cluster which is a standard configurable component of SPARC SuperCluster.
  • And in terms of further reducing risk, this solution is based on SPARC SuperCluster, which has fully redundant components, from compute nodes to networking, and also offers built-in high availability and fail over with Solaris Cluster.During our testing, we measured and documented all the critical failure detection and recovery times. For example a failure of the Web tier was detected by Oracle Solaris Cluster in 13 seconds and the fail over to the remaining node completed in 11 seconds. A complete recovery of the failed Web container back to full redundancy only took 68 seconds. A failure of the App tier was detected in 6 seconds and failed over in 4 seconds.
  • You can start planning your path right nowArrange for an onsite tech dayRisk assessmentTrainingOracle Platinum ServicesZDnet: Oracle aims to redefine IT support“[remote telemetry and response]” without loss of controlChoose your options for de-risking your data centerOptimized Solutions, Engineered Systems,building block components
  • Transcript

    • 1. 1 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 2. Creating a High AvailabilityArchitecture with SPARCSuperClusterCarter WrightDatabase Engineer, Computer Knowledge, Inc.Bruce ChapmanPrincipal Software Engineer, OracleDean HalbeisenSolutions Manager, Oracle2 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 3. Program  SPARC SuperCluster High AvailabilityAgenda Implementation  SPARC SuperCluster High Availability  Optimized Solutions High Availability Examples  Next Steps / Resources3 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 4. SPARC SuperCluster High AvailabilityLarge Government Law Enforcement Agency Implementation of HA Leverage the HA features built in to the SPARC SuperCluster – Oracle Real Application Cluster(RAC), Oracle Exadata Storage Server, Oracle Automatic Storage Management(ASM), Network and Power redundancies – Clustering of applications in Oracle VM Server For SPARC and Oracle Solaris Zones on SPARC T4-4 nodes Use of Oracle Data Guard Use of Oracle RMAN Use of Sun ZFS Storage Appliance 7420 and ZFS Replication4 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 5. SPARC SuperCluster High AvailabilityWorkload Overview 12,000 to 15,000 concurrent users worldwide during peak hours Database greater than 1.6TB and growing 1500 to 2000 SQL executions per second per RAC node during peak hours (~4000 SQL execs per second combined) Greater than 1.5 million blocks per second read from each SGA during peak hours ( > 3 million blocks per second combined ) 900K to 1.2MB of redo written per second (combined RAC nodes) Query response time typically less than 300ms Data Guard latency less than three seconds to standby site ZFS replication latency less than three seconds to standby site Each RAC node CPU utilization at peak load 15% to 20%5 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 6. SPARC SuperCluster High AvailabilitySPARC SuperCluster Configuration Solaris 11 Solaris 10 Nodes 1 & 2 AppExternal ZoneServer1 10GbE Solaris 10 ZFS 7420 Storage Nodes 3 & 4 InfiniBand RMAN Backups Data Pump DirExternal 600GServer2 App Svr 10K SAS App App Zone Zone InfiniBand Exadata Data CellsExternal 216TB RawServer3 ASM High 72TB Usable 3 X ASM Disk Groups6 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 7. SPARC SuperCluster High AvailabilityDatabase Configuration SuperCluster installation does provide a fully functional Oracle RAC install, but it has to be configured to meet your applications needs – Database Hardening must be performed – Listener configurations (scan and local) can be modified – You configure your RAC Workload Services – Network settings – 10Gb and 1Gb interfaces only  TCP window, MTU’s  Leave IB settings alone – Backup and recovery is your responsibility – Application tuning is your responsibility7 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 8. SPARC SuperCluster High AvailabilityDatabase Configuration – OLTP or Data Warehouse (or both) – SGA Settings can be modified  We changed only a few settings – SGA, PGA Aggregate Target – Data Guard parameters  SPARC SuperCluster uses Auto SGA and Auto PGA, not full AMM (see pg 18-18 11g Concepts Guide)  Can change the quantity and size of Redo Logs (if needed)  Users, tablespaces, datafiles – Roles and privileges – Datafile size and quantity8 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 9. SPARC SuperCluster High AvailabilityWorkload Services Intranet SCAN + Service Name SCAN Listeners Load Balanced Services Single Node Services Main App 1 Main App 2 Smaller Apps Non-RAC COTS RMAN Service Data Guard Data Guard 12K-15K Active Concurrent Users Accessed World Wide9 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 10. SPARC SuperCluster High AvailabilityData Guard on SPARC SuperCluster  Data Guard setup – Physical Standby, High Performance Mode, Async  Multi-instance Primary, single instance standby  Active Guard enabled at standby site  Use standby for – Reporting and Read-only access – Smoke testing application deployments  Make standby a temporary Primary and restore back to standby using Flashback Database  Use standby as the Last deployment before deployment to prod site – Other live applications to leverage investment10 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 11. SPARC SuperCluster High AvailabilityReplication Strategy Primary Standby SPARC Data Guard Replication SPARC SuperCluster Database Content SuperCluster Latency to Remote Site <3 Seconds ZFS Storage ZFS Replication ZFS Storage Appliance 7420 External Content Appliance 7420 Latency to Remote Site <3 Seconds11 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 12. SPARC SuperCluster High AvailabilityRMAN Configuration  Backupsets stored on ZFS 7420 (not in ASM) Solaris 11 Solaris 10 Node 2 – Accessible during crisis App RMAN Services – Leave space to database in Exadata Zone – Backup file system mounted to both nodes for parallelism – ZFS RAID 10 Solaris 11 Solaris 10  Use RAC Workload Service Node 1 App – Allows RMAN parallelism across RAC Zone  rman target sys/<pw>@rman_svc  Improves speed ZFS 7420 Storage InfiniBand  Parallelism 16 with 16 Channels allocated RMAN Backups – RMAN will allocate 8 on node 1, 8 on node212 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 13. SPARC SuperCluster High AvailabilityRMAN Backup Operations  Level 0 weekly – 7 day recovery window  Level 1 nightly – Cumulative Incremental  Backup both primary and standby sites separately13 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 14. SPARC SuperCluster High AvailabilityRMAN Configuration and Management in a RAC Environment  Recovery Catalog is required  Flashback database enabled at both sites  db_unique_name required  Backups taken at both sites – BACKUP and RESTORE command uses TAG ‘system_name’ option when creating backups  Distinguishes each site in Recovery Catalog  Archivelog deletion policy on Primary – Ensures archivelogs are not deleted by RMAN before getting to standby site CONFIGURE ARCHIVELOG DELETION POLICY TO SHIPPED TO ALL STANDBY;14 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 15. SPARC SuperCluster High AvailabilityMonitoring and Management  Grid Control 12c – Good graphs and drill down capabilities  Custom Scripts (bash) – Exadata Cell monitoring –  No Grid 12c plug-in yet for SPARC SuperCluster with Exadata – Alert monitoring to send notifications to our NOC15 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 16. SPARC SuperCluster High AvailabilityTime to Deployment for an Engineered Solution  Early adopter of SPARC SuperCluster and it was not your common install being a customer engineered application, had to train a new team with clearance to work on the systems to perform the install  Purchased three SPARC SuperCluster systems in December 2011 Pre-Prod, Production, Standby DR site  First system delivered early February, next system two weeks later, third system two weeks after that  Each system was installed and running our application within three weeks of each delivery.  Within a 9 week period, we had three fully operational SuperClusters in three different data centers.  Went live production in July of 2012. Successful worldwide launch  Instantiated DR site with Oracle Data Guard shortly thereafter.16 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 17. SPARC SuperCluster High AvailabilityKnowledge to Share  SPARC SuperCluster High Availability White papers on www.cknowledge.com – How To Guide: RAC Workload Services – 11gR2 – How To Guide: Building Data Guard – RMAN Quick Reference17 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 18. Program  SPARC SuperCluster High AvailabilityAgenda Implementation  SPARC SuperCluster High Availability  Optimized Solutions High Availability Examples  Next Steps / Resources18 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 19. SPARC SuperCluster: High AvailabilityHardware HA Features Mirrored DisksDual Power Distribution Units All OS Disks On T4-4 Domains areBoth connected to each server or switch mirrored.in the SuperCluster, each with redundantpower supplies. SPARC T4-4 HA FeaturesTen Gigabit Ethernet Networking Hot swap redundant PS, hot swap PCI-E, DIMM off lining, predictive selfEach logical 10GbE interface has two healingphysical ports for connection to separateexternal lab switches. 7320 ZFS Storage Appliance -Two Storage head nodes each withInfiniband Networking redundant networking, SCSI.- 3 IB switches for availability / growth. -Configurable RAID / spare options for- Each logical IB connection to T4-4, shares.Exadata DB Storage Cells, and 7320 Database Storage CellsStorage Appliance has two physical Each with redundant IB andInfiniBand ports. configurable disk mirroring.19 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 20. SPARC SuperCluster: High Availability Software HA Multiple RAC instances Enabling DB to tolerate and react to instance and node failures automatically with zero to near zero impact on applications. Oracle Grid Infrastructure Clusterware and ASM provide abstraction of DB storage cell nodes and disks, also able to continue functioning despite disk or DB storage node failure. Oracle Solaris Cluster One of several high availability options for application domains. Configurable number of Domains for horizontal scaling and failover.20 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 21. Program  SPARC SuperCluster High AvailabilityAgenda Implementation  SPARC SuperCluster High Availability  Optimized Solutions High Availability Examples  Next Steps / Resources21 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 22. High Availability on SPARC SuperCluster Oracle Optimized Solutions 10 GbE Data Center Network T4-4 Node 3 T4-4 Node 4 T4-4 Node 1 T4-4 Node 22 T4-4 Node Oracle Solaris Oracle Oracle Solaris Oracle Solaris Solaris Oracle Solaris Cluster High GP Domain GP Domain GP Domain Availability GP Domain Oracle Solaris 11 Oracle Oracle Solaris 11 RAC Oracle Solaris 11 11g R2 Oracle Solaris 11 Scale Database Domain Database Domain Database Domain Database Domain InfiniBand Network Oracle Exadata Storage Servers22 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 23. High Availability / Fail Over Test Results Extremely fast failure detection and service recovery times Detection Recovery Web server node failure 11 12 OPMN/OC4J failure 0 0 SECONDS General purpose domain failure 0 10 Database domain failure 61 523 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 24. Driving Risk Down • Driving Innovation Up • Active Partnership with OracleUp • Active Partnership with Oracle • Driving Risk Down • Driving InnovationDown • Driving Innovation Up • Active Partnership with Oracle • Driving Risk Driving Risk Down • Driving Innovation Up • Active Partnership with OracleUp • Active Partnership with Oracle • Driving Risk Down • Driving InnovationDown • Driving Innovation Up • Active Partnership with Oracle • Driving Risk Tested, Secure, Flexible Driving Risk Down • Driving Innovation Up Risk Down Driving • Active Partnership with OracleUp • Active Partnership with Oracle • Driving Risk Down • Driving Innovation Driving Innovation UpDown • Driving Innovation Up • Active Partnership with Oracle • Driving Risk Driving Risk Down • Driving Innovation Up • Active Partnership with Oracle Active Partnership with OracleUp • Active Partnership with Oracle • Driving Risk Down • Driving InnovationDown • Driving Innovation Up • Active Partnership with Oracle • Driving Risk Driving Risk Down • Driving Innovation Up • Active Partnership with Oracle 24 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 25. Program  SPARC SuperCluster High AvailabilityAgenda Implementation  SPARC SuperCluster High Availability  Optimized Solutions High Availability Examples  Next Steps / Resources25 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 26. SPARC SuperCluster Conference Sessions Session Date Time RoomAccelerate Oracle E-Business Suite Deployment with SPARC SuperCluster Monday 12:15PM Moscone West - 2001SPARC SuperCluster: Architecture, Roadmap, and Partner Services Opportunities Monday 12:15PM Marriott Marquis – Golden Gate BMy Oracle Support: Tips and Techniques for Getting the Best Hardware Support Monday 1:45PM Moscone West - 2001Top 10 Database Performance Tips for SPARC Systems Running Oracle Solaris Monday 1:45PM Moscone West - 2000Building a Flexible Enterprise Cloud Infrastructure on Oracle SPARC Systems Monday 4:45PM Moscone West - 2001Database and Application Consolidation on SPARC SuperCluster Monday 4:45PM Moscone West - 2000Oracles SPARC Server Strategy Update Tuesday 10:15AM Moscone West - 2001Oracle Engineered Systems Backup and Recovery Best Practices Tuesday 10:15AM Moscone South - 252Transform Data Center TCO with Oracle Optimized Servers: A Customer Panel Tuesday 11:45AM Moscone West - 2001Deep Dive into SPARC SuperCluster: Engineered for Extreme Efficiency Tuesday 1:15PM Moscone West - 2001Accelerate Your SAP Landscape with the Oracle SPARC SuperCluster Tuesday 5:00PM Moscone West - 2001High Performance Security for Oracle Applications Using SPARC T4 Systems Tuesday 5:00PM Moscone West - 2000SPARC Processor Update Wednesday 10:15AM Moscone West - 2000Creating a Maximum Availability Architecture with SPARC SuperCluster Wednesday 11:45AM Moscone West - 2000Improve PeopleSoft HCM Performance and Reliability with SPARC SuperCluster Wednesday 5:00PM Moscone South - 200Managing Sun Servers and Oracle Engineered Systems with Oracle Enterprise Manager Thursday 12:45PM Moscone West - 2000 26 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 27. SPARC SuperCluster Demos Demo LocationAccelerating SAP Queries and Reporting on SPARC SuperCluster Moscone South -102Bullet Proof E-Business Suite on SPARC SuperCluster Moscone South -101Bullet proof PeopleSoft on SPARC SuperCluster Moscone South-131Consolidate your Datacenter: Enterprise Cloud Infrastructure on SPARC Moscone South-150Explore the Benefits of Oracle SPARC SuperCluster Moscone South-104Mission Critical Availability and Disaster Recover for Your Cloud Moscone South-116Mission Critical SPARC-Powered Infrastructure as a Service Moscone South-157Oracle SPARC SuperCluster: World’s First General Purpose Engineered System Moscone South-106Oracle SPARC SuperCluster T4-4 to Improve Efficiency and Cut Costs Moscone South-105The Data Center Superhero: Backup and Recovery Systems Moscone South-107Transform an Enterprise Cloud Infrastructure: Oracle SPARC SuperCluster Moscone South-130Transform and Optimize Revenue Management with SPARC SuperCluster Moscone South-108Engineered System Showcase - SPARC SuperCluster Consolidation Moscone North Lower Lobby 27 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 28. Resources In depth information  Computer Knowledge White Papers – http://www.cknowledge.com  Oracle Maximum Availability Architecture Guides – http://www.oracle.com/technetwork/database/features/availability/maa- 090890.html  Optimized Solutions – http://www.oracle.com/us/solutions/oos/overview/index.html28 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 29. Graphic Section Divider29 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 30. 30 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.
    • 31. 31 Copyright © 2012, Oracle and/or its affiliates. All rights reserved.