HP & Oracle A Powerful Alliance Bei Fragen wenden Sie sich bitte an: Frau Virpi Matikainen Oracle Account Managerin email: virpi_matikainen@hp.com Tel: 089/61412-513 Rebecca Kühn HP/Oracle CTC 10. Dezember 2001 filename\location Page 1
hp/oracle alliance strong and growing oracle is hp's #1 database partner hp is tier1 platform for oracle hp is oracle's #1 server platform partner hp is oracle's #1 applications partner more than 200 professionals are dedicated to the alliance filename\location Page 2
hp/oracle alliance HP/Oracle test configurations: core development is done on hp systems CRM releases are already first on hp joint product development, optimization and IA64 product collaboration filename\location Page 3
Oracle s 100% commitment to hp Oracle operates ww consolidated email systems on HP-UX clusters entire Oracle email infrastructure runs on 2 HP V-Class servers implemented in 90 days 40,000+ users reduced costs ERP and CRM being migrated now entire e-business runs exclusively by 2001 filename\location Page 4
hp/oracle cooperative technology center located at hp in Böblingen and Oracle UK founded in spring 1994 oracle and hp people in one team part of hp/oracle alliance EMEA filename\location Page 5
Data Server CTC Focus: HP/Oracle CTC Oracle Databases 7.x, 8.x, 9i (32/64-bit) High Availability (OPS / RAC / OPFS / Cluster / HA) Data Warehousing / Business Intelligence Middleware CTC Oracle Applications CTC Oracle Internet Application S<erver (ias) Manageability (HP OpenView / Oracle Enterprise Manager) Oracle Applications ERP Oracle Customer Relationship Management (CRM) ctc_update101100.ppt Page 6
hp/oracle ctc activities technology consulting for European partners and customers and sales teams know-how transfer to EMEA local organizations evaluation and tests of new products/features technology transfer to and from US labs presales technical assistance (performance tests, tuning & analysis, demo, sizing, benchmarking) filename\location Page 7
Integrated Products Oracle Enterprise Manager OEM Value Add Packs Certified Configurations Oracle CRM Oracle E-Business Suite 11i 9i Real Application Clusters Oracle9i Application Server Oracle9i Recovery Manager Oracle9i Database HP OpenView Operations HP OV Network Node Manager HP OpenView OmniBack II HP Cluster Manager HP ServiceGuard OPS edition HP-UX11i HP NetServer, IA-64 HP XP512/ FC60 / Virtual Array HP 9000 Server HP/Oracle Integration Technology Stack ctc_update101100.ppt Page 8
The objectives of the ETC are: Ensure our joint customers and partners are successful in implementing highly scalable and mission critical solutions Focus on Mission Critical Computing: 9i RAC on HP9000 Servers, XP-Storage Systems, e-services/e-business Applications Showcasing the latest Oracle solutions on HP to our partners and customers filename\location Page 9
Oracle9i Handles Causes of Downtime Unplanned Downtime System Failure Data Failure & Disaster Human Error Real Application Clusters Fast Restart Recovery Manager Data Guard Flashback Query Log Miner, Data Guard Planned Downtime System Maintenance Database Maintenance Dynamic Reconfiguration Data Guard, RAC Online Redefinition Partitioning, Parallel SQL ctc_update101100.ppt Page 10
Traditional High Availability Solutions Today, applications with high availability service requirements using an Oracle database are typically deployed using one of two architectures: Single Instance Oracle with MC/ServiceGuard Oracle Parallel Server (OPS) / Oracle9i Real Application Cluster (9i RAC) with HP ServiceGuard OPS Edition filename\location Page 11
MC/ServiceGuard Non-Shared Database Provides 16 node failover solution Virtual Server with virtual IP Address Before Failover: Virtual IP Adress and Networkname refers to System A After Failover: Virtual IP Adress and Networkname refers to System B Oracle SystemA Disk A Oracle SystemB Disk B ctc_update101100.ppt Page 12
MC/ServiceGuard Node 1 Pkg A Pkg B Pkg C Node 2 Pkg D Pkg E Pkg A Node 4 Balance workload after a node failure Minimize impact on remaining nodes Pkg B Node 3 Pkg F Pkg G If Node 1 fails... Pkg C Pkg H Pkg I ctc_update101100.ppt Page 13
Single Instance Oracle with MC/ServiceGuard Active Oracle Instance 1 failover Active Application 2 HP MC/ServiceGuard HP-UX 10.20 or HP-UX 11.x Exclusive Access Database for Oracle Instance 1 HP MC/ServiceGuard HP-UX 10.20 or HP-UX 11.x Exclusive Access Application 2 Storage ctc_update101100.ppt Page 14
Limitations of Cold Failover (MC/ServiceGuard) Scalability of cluster is limited to scalability of one server Cold failover is slow, as many time consuming tasks must be performed as part of failover moving and mounting logical volumes starting the oracle instance opening the data files After failover, all sessions must re-establish connections to the database After failover, the instance caches are cold introducing a performance brownout ctc_update101100.ppt Page 15
Improves upon cold failover 8i Oracle Parallel Server/ 9i Real Application Clusters Increase Scalability beyond one system Much faster failover and switchover no need to move volumes or restart the Oracle instance Can preconnect users to the secondary instance for fast restart without logon Caches are warm on the secondary instance -- better performance after failure ctc_update101100.ppt Page 16
8i OPS / 9i RAC Architecture High-speed bus Node 1 Node 2 Node 3 Memory Memory Memory CPUs CPUs CPUs shared HP XP Storage System ctc_update101100.ppt Page 17
8i OPS / 9i RAC Software Stack Active Active Oracle 9i RAC Instance 1 Cluster Interconnect Oracle 9i RAC Instance 2 HP ServiceGuard OPS Edition HP ServiceGuard OPS Edition HP-UX 11.x HP-UX 11.x Shared Access Shared Database for Oracle Instances 1 and 2 Shared Access ctc_update101100.ppt Page 18
Problem: Stale Read Node 1, Instance A SGA USER 3 6??? IC Node 2, Instance B SGA B1 FRED USER 1 USER 2 B1 FRED 2 HUGO 4 B1 FRED HUGO BOB 1 LGWR 3 5 B1 FRED Data A-Z Redo- Logs Shared Disk Subsystem ctc_update101100.ppt
Cache Coherency Protocol: Ping Node 1, Instance A 1 GCM Node 2, Instance B SGA USER 3 6 SGA B1 BOB???!!!! 5 2 B1 FRED HUGO BOB 7 IC DBWR LGWR 4 3 B1 Shared Disk Subsystem BOB Data A-Z Redo- Logs ctc_update101100.ppt
Cache Fusion Optimization of Cache Coherency Protocol Fewer messages Avoid I/O Cache Fusion provides copies of blocks directly from holding instance s memory cache to the requesting instance s memory cache. Implementation in two phases Oracle8i Parallel Server, Cache Fusion I Oracle9i Real Application Clusters Cache Fusion II filename\location Page 21
Cache Fusion Manages Inter Instance Block Requests Request for Block Cache A Read Read Write Write Lock Status Block in Cache B Read Write Read Write 8i, Cache Fusion Phase I 9i, Cache Fusion Phase II ctc_update101100.ppt Page 22
Cache Coherency Protocol: Cache Fusion Node 1, Instance A 1 GCM Node 2, Instance B SGA USER 3 2 SGA B1 BOB 3 Block Shipping B1 FRED HUGO BOB IC DBWR LGWR Shared Disk Subsystem B1 FRED Data A-Z Redo- Logs ctc_update101100.ppt
Cache Fusion Performance Without Cache Fusion 100 msec Cache Fusion Block access time 20 msec 1 msec.01 msec Block in local cache Block in remote cache Block on disk ctc_update101100.ppt Page 24
RAC Test Results Up to 72 continuous hours of cluster operations under stress Successful cluster reconfiguration due to destructive tests Concurrent usage of XP512 by HP-UX and Windows 2000 200+ individual product improvements 1.8 scalability with 2-node N4000 cluster 3.5 scalability with 4-node N4000 cluster 6.4 scalability with 8-nodes N4000 cluster filename\location Page 25
HP HyperFabric: The HP 9000 Virtual Backplane 2-8 Server Engines, Point-to-Point 2-64 Server Engines, Switched Configuration Fast and flexible data center and multi-system application interconnect Super-high bandwidth (2.5 Gb/s per card) Low-latency via reduced TCP/IP stack at <50 µsec Mixed HP 9000 servers (L-, N-, V-Class, SuperDome and future HP 9000 servers) Mixed HP-UX version support (HP-UX 11i, 11.0, 32-/64-bit) 10km switch-to-switch distance filename\location Page 26
Bandwidth is increased to 4 Gb/s Compatible with HyperFabric1 HyperFabric 2 improvements for Oracle9i Cache Fusion available on IA and PA servers & workstations supports OLAR and DLKM of IO cards Hyper Messaging Protocol (HMP) first implementation of low latency, high bandwidth protocol with 20% performance improvement over UDP Co-developed by HP and Oracle for exemplary performance Lower latency to 22-25 µsec with hyper messaging protocol (HMP)
Active-Active Mode HP HyperFabric : Highly Available Simultaneously active links Service interruption causes HyperFabric driver to move traffic off of failed link, and onto a good link Redundant Switch Mode TCP/IP session stays up (unlike traditional networking technologies) Two switches are set up with redundant connections to each engine Certified with HP ServiceGuard and Oracle9i RAC filename\location Page 28
hp rp5400 series introducing the world s first clustered Oracle Application Standard Benchmark Release 11i
world s first clustered Oracle Application Standard Benchmark Release11i four rp5470 servers 1.25 seconds 4,368 users two rp5470 servers 1.16 seconds 2,296 users breaking even more new ground with near perfect linear scalability!
the secrets of our success industry leading rp5470 server performance accelerated performance with HP s Hyper Messaging Protocol (HMP) 5+ years of joint solutions development and optimization
a look at the configurations two rp5470 servers with 4 PA- 8600 550MHz processors, 16GB RAM, 2 X 36GB internal disks, FC60 with 5.36TB, HP-UX 11.0 HP Hyper Messaging Protocol over HyperFabric II Oracle Application Standard Benchmark Suite, Release 11i Oracle9i database with Real Application Clusters (RAC) 2,296 users! four rp5470 servers with 4 PA- 8600 550MHz processors, 16GB RAM, 2 X 36GB internal disks, FC60 with 5.36TB, HP-UX 11.0 HP Hyper Messaging Protocol over HyperFabric II Oracle Application Standard Benchmark Suite, Release 11i Oracle9i database with Real Application Clusters (RAC) 4,368 users! ctc_update101100.ppt Page 32
first clustered Oracle Application Standard Benchmark Release 11i Oracle Application Standard Benchmark Release 11i: some background information measures performance and scalability of Oracle applications on specific system configurations simulates realistic customer scenarios using a selection of the most commonly used Oracle applications modules utilizes Oracle9i database ctc_update101100.ppt Page 33
Oracle9i RAC: Transparent Application Failover Oracle9i provides a high availability architecture that provides transparent client failover capability Little or no user downtime Applications and users are automatically and transparently reconnected to another system Applications and queries continue uninterrupted Login context maintained Computer A Computer B Node A in an RAC cluster fails, users are migrated Computer A Computer B
TAF Select Failover Failover allows the application to continue execution or fetching Leverages Oracle s multi-versioning read consistency to ensure results are identical SELECT * FROM emp; Client SELECT * FROM emp; empno name empno name 7369 Smith 7499 Allen 7521 Ward 7566 Jones 7654 Martin 7698 Blake Rows Partially Returned When Failover Occurred Instance 1 DB Instance 2 7369 Smith 7499 Allen Continues 7521 Returning Ward Remaining 7566 Rows Jones from here 7654 Martin 7698 Blake ctc_update101100.ppt Page 35
Oracle9i Best Fit - MC/ServiceGuard vs. Oracle9i RAC MC/ServiceGuard Minor downtime acceptable Lower cost of downtime Scalability beyond one system not required Raw Devices or File Systems Any Oracle-based applications Oracle9i RAC High Availability is mission critical High Cost of Downtime Scalability beyond one system essential Raw Devices Oracle8i: Partitioning - Application or data ctc_update101100.ppt
Oracle8i Parallel Server/ Oracle9i RAC Oracle Parallel Server can be optimized for high availability It can be integrated with the technology of the systems vendors to provide even higher HA Notify client sessions of failover Eliminate time-outs by moving IP addresses Monitor external to Oracle Make other applications on cluster HA You could do this, but. filename\location Page 37
Oracle8i Parallel Fail Safe / Oracle9i RACGuard Let Oracle manage this complexity Oracle Parallel Fail Safe is a prepackaged, pre-configured, pre-tested high availability solution Built upon 9i RAC (8i OPS) Integrated with high availability products from HP Compatible with off-the-shelf and custom applications Formerly a separate product, now an integrated feature of Oracle9i Real Application Clusters filename\location Page 38
8i OPFS / 9i RACGuard Special Primary/Secondary Parallel Server configuration optimized for High Availability Application services provided by the primary instance Secondary instance ready to take over in the event of a failure Tightly integrated with cluster manager for fast detection and response. Recovers quickly after failover because it s Parallel Server ctc_update101100.ppt Page 39
Additional HA optimizations 8i OPFS / 9i RACGuard Monitor complete Oracle environment to ensure clients can connect and do useful work Automatically recover from failures external to the database that affect database access Incorporate TAF to automatically reconnect or preconnect sessions Intelligently capture diagnostics information after a failure Reduce time to interrupt the client session Reduce reconnect time due to networking delays Pre-warm the library cache on the secondary instance. ctc_update101100.ppt Page 40
8i OPFS / 9i RACGuard Operation A) Single Database on a cluster Primary IP Primary Instance A Secondary IP Secondary Instance A Server 1 Server 2 Database A ctc_update101100.ppt Page 41
8i OPFS / 9i RACGuard Operation (cont.) A) Single Database on a cluster: Failure of Server1 Primary IP Primary Instance A Secondary IP Secondary Instance A Server 1 Server 2 Database A ctc_update101100.ppt Page 42
8i OPFS / 9i RACGuard Operation (cont.) B) Multiple Databases on a cluster Primary IP Secondary IP Secondary IP Primary IP Primary Instance A Secondary Instance B Secondary Instance A Primary Instance B Server 1 Server 2 Database A Database B ctc_update101100.ppt Page 43
8i OPFS / 9i RACGuard with ServiceGuard OPS Edition Primary/Active Secondary/Standby Oracle9i RAC Oracle HA Packs (Instance,Access) HP ServiceGuard OPS Edition Interconnect DLM System Management Infrastructure Oracle9i RAC Oracle HA Packs (Instance,Access) HP ServiceGuard OPS Edition K, N or V class HP-UX 11.x K, N or V class HP-UX 11.x Shared Access Shared Database for Oracle Instances 1 and 2 Shared Access ctc_update101100.ppt Page 44
9i RACGuard Recovery RACGuard solution for the highest database availability needs Exact failover times dependent on configuration, database size and other variables Recovery Time Estimates Parallel Fail Safe Solution 1-3 minutes Oracle 8i with MC/ServiceGuard 4-10 minutes Oracle 8 with MC/ServiceGuard 10-60 minutes ctc_update101100.ppt Page 45
Failure Detection Time Failure Class Detection Process Detection Time Database Failure Database Hang System Failure Oracle process monitor detects missing processes and initiates failover Monitor detects and halts the instance. Failover is initiated ServiceGuard OPS Edition detects failure, reconfigures the cluster, and notifies Oracle 10 seconds Configurable depending upon tolerance for false failover. Typically 200 seconds 30-60 seconds (using cluster lock disk) ctc_update101100.ppt Page 46
Failover Time Failover Step Group Membership Services reconfigures Distributed Lock Manager reconfigures Failover Time: Oracle Parallel Fail Safe 10 seconds n/a 5 seconds n/a Failover Time:Failover without Oracle Parallel Fail Safe Volumes Fail Over n/a 10 to 20 seconds Restart Oracle n/a Up to 5 minutes Instance Recovery 20 seconds Up to 30 minutes with Oracle 8 ctc_update101100.ppt Page 47
Oracle9i Best Fit - Oracle9i RAC vs. Oracle9i RAC Guard Oracle9i RAC Oracle9i RAC Guard High cost of downtime Raw devices High Availability required Scalability beyond one system essential Highest levels of performance a requirement Only detect node or instance failure Highest levels of availability required Scalability beyond one system not required Some performance trade-offs Detect hanging instance/ performance degradation ctc_update101100.ppt
HP/Oracle test configurations: Thank You! filename\location Page 49