2 Why Oracle Software Runs Best on Oracle Hardware Yufei Zhu, Manager PAELily Zhao, Senior Engineer
3 Program Agenda Oracle Software Performance and SPARC T5 Oracle Engineered System’s ApproachTechnology Areas CoveredConclusions
4 The following is intended to outline our general product direction 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.
6 T5-8 TPC-C Benchmark: OLTP DB Leadership World Record OLTP single-system performanceServerso,coreProcessorSWtpmC$/tpmCOracle SSC27n x 4chSPARC T3Oracle 11g30.2M$1.01IBM Power 7803n x 8ch3.86 Power7DB2 9.710.3M$1.38Oracle T5-88, 1283.6 SPARC T58.55M$0.55IBM p59532, 645.0 Power6DB2 9.56.0M$2.81Oracle X48008, 802.4 Xeon E75.0M$0.94Cisco UCS2, 162.9 E5-26901.6M$0.472, 84.14 Power71.2M$0.69World record single-system TPC-C 8,552,523 tpmCSPARC T5-8 delivers 3.1M IOPs from storageSPARC T5 2.4x faster per chip performance than IBM configuration3-node IBM p780 (24 total chips at 10.3M), Oracle has 2.5x better $/tpmCSPARC T5 5x better response time than IBM p780 cluster (New Order 90%)Oracle's SPARC T5-8 server equipped with eight 3.6 GHz SPARC T5 processors obtained a result of 8,552,523 tpmC on the TPC-C benchmark. This result is a world record for single servers. Oracle demonstrated this world record database performance running Oracle Database 11g Release 2 Enterprise Edition with Partitioning.The SPARC T5-8 delivered a single system TPC-C world record of 8,552,523 tpmC with a price performance of $0.55/tpmC using Oracle Database 11g Release 2. This configuration is available 09/25/13.The SPARC T5-8 has 2.8x times better performance than the 4-processor IBM x3850 X5 system equipped with Intel Xeon processors.The SPARC T5-8 delivers 1.7x the performance compared to the next best 8 chip result.The SPARC T5-8 delivers 2.4x the performance per chip vs. IBM Power node cluster result.The SPARC T5-8 delivers 1.8x the performance per chip vs. IBM Power 780 non-clustered.The SPARC T5-8 delivers 1.4x the performance per chip compared to the IBM Flex x240 Xeon result.The SPARC T5-8 delivers 1.7x the performance per chip compared to the Sun Server X2-8 system equipped with Intel Xeon processors.The SPARC T5-8 demonstrated over 3.1 Million 4KB IOP/s with 76% idle, demonstrating its ability process a large IO workload with lots of processing headroom.This result showed Oracle's integrated hardware and software stacks provide industry leading performance.The Oracle solution utilized Oracle Solaris 11.1 with Oracle Database 11g Enterprise Edition with Partitioning and demonstrates stability and performance with this highly secure operating environment to produce the world record TPC-C benchmark performance.See benchmark disclosure slide
7 T5-8 SPECjEnterprise: Enterprise Java Leadership World Record performance SPECjEnterprise2010App Serverchip,coreProcessorEjOPsSWOracle T5-88, 1283.6 SPARC T557,422.17WLS 12.1Oracle 4x T4-416, 1283.0 SPARC T440,104.86WLS 10.3Oracle X4800 M28, 802.4GHz E7-887027,150.05IBM Power c8, 643.86 Power716,646.34WS 7.0IBM PL7R2 8c2, 164.2 Power7+12,062.46WS 8.5IBM X36502.9 E5-26909,696.43SPARC T5-8 application & T5-8 database achieved 57, EjOPsT x better App Tier $/perf: Power 780 = $992k, T5-8 = $269kT x faster than 8-chip IBM Power 780, 1.7x faster per core than Power7SPARC T x faster than 8-chip x86 Xeon E7-8870SPARC T5-8 DB server only 40% utilized, headroom to spareOracle produced a world record SPECjEnterprise2010 benchmark result of 57, SPECjEnterprise2010 EjOPS using Oracle's SPARC T5-8 server in the application tier and another SPARC T5-8 server for the database tier.The SPARC T5-8 server demonstrated 3.4x better performance compared to an 8-socket IBM Power 780 server result of 16, SPECjEnterprise2010 EjOPS. The SPARC T5-8 is 3.7x less expensive for the application server hardware list cost compared to the IBM configuration.The SPARC T5 processor demonstrated 1.7x better performance per core compared to the POWER7 processor used in the IBM Power 780 SPECjEnterprise2010 result.The SPARC T5-8 server demonstrated 2.2x better performance compared to the Cisco UCS B440 M2 Blade Server result of 26, SPECjEnterprise2010 EjOPS.The SPARC T5-8 servers used in the application and database tiers ran the Oracle Solaris 11.1 operating system.The SPARC T5-8 server for the application tier used Oracle Solaris Zones to consolidate sixteen Oracle WebLogic Server instances to achieve this result.The SPARC T5-8 server for the database tier was 40% utilized during the execution of the benchmark allowing additional room for growth.This result demonstrated less than 1 second response time for all SPECjEnterprise2010 transactions, while demonstrating a sustained load of Java EE 5 transactions equivalent to 468,000 users.A single SPARC T5-8 server in the database tier producing 57, SPECjEnterprise2010 EjOPS at 40% utilization can replace a total of 2x SPARC T4-4 servers that obtained 40, SPECjEnterprise2010 EjOPS. A single SPARC T5-8 server in the database tier producing 57, SPECjEnterprise2010 EjOPS at 40% utilization can replace 6x SPARC T3-2 servers where each SPARC T3-2 server obtained 9, SPECjEnterprise2010 EjOPSSee benchmark disclosure slide
8 T5-8 SPECjEnterprise: Enterprise Java Leadership World Record virtualized (1 server: 1VM for App & 1VM for DB)App + DB Serverchip,coreProcessorEjOPsSWOracle T5-8, Zones8, 1283.6 SPARC T527,843.57WLS 12.1IBM Power 780, LPAR8, 324.42 Power7+10,902.30WS 8.5SPARC T5-8 virtualized at 27, EjOPsSPARC T5 is 2.6x faster than IBM’s Power7+Oracle’s price/performance advantageSPARC T x better server $/perfPower 780 = $835k, T5-8 = $297kSPARC T x better HW+SW $/perf than 780Also beats PowerLinux 7R2 $/perf (2x 2-chips)Disproves that per-core performance reduces costOracle produced a world record single-server SPECjEnterprise2010 benchmark result of 27, SPECjEnterprise2010 EjOPS using one of Oracle's SPARC T5-8s for both the application and the database tier. This result directly compares the 8-chip SPARC T5-8 (8 SPARC T5s) to the 8-chip IBM Power 780 (8 POWER7+). The 8-chip SPARC T5 is 2.6x faster than the 8-chip IBM POWER7+ server.Both Oracle &IBM used virtualization to provide 4-chips for application and 4-chips for DB.The server cost/performance for the SPARC T5 server was 7.1x better than the IBM POWER7+. The cost/performance of the SPARC T5-8 is $10.72 compared to the IBM Power 780 at $76.64.The total configuration cost/performance (hardware+software) for the SPARC T5 was 3.6x better than the IBM POWER7+. The cost/performance of the SPARC T5-8 is $56.21 compared to the IBM Power 780 at $ The IBM system had 1.6x better performance per core, but this did not reduce the total SW+HW. This shows performance-per-core is a poor predictor of characteristics relevant to customers.IBM has a non-virtualized result (1server for application &1server for database). The IBM PowerLinux 7R2 achieved 13, SPECjEnterprise2010 EjOPS which means it was 2.1x slower than the SPARC T5-8 server. The total configuration cost/performance (HW+SW) for the SPARC T5 server was 11% better than the IBM POWER7+ server. The cost/performance of the SPARC T5-8 is $56.21 compared to the IBM PowerLinux 7R2 at $62.26.The total IBM hardware plus software cost was $2,174,152 versus the total HW+SW cost of $1,565,092. IBM could only provide 768 GB of memory while Oracle was able to deliver 2 TB in the SPARC T5-8.The SPARC T5-8 requires only 8 rack units, the same as the space of the IBM Power IBM had a hardware core density of 4 cores per rack unit which contrasts with the 16 cores per rack unit for the SPARC T5-8. Again performance-per-core is a poor predictor of characteristics relevant to customers.The virtualized SPARC T5 server ran the application tier servers on 4 chips using Oracle Solaris Zones and the database tier in a 4-chip Oracle Solaris Zone. The virtualized IBM POWER7+ server ran the application in a 4-chip LPAR and the database in a 4-chip LPAR.The SPARC T5-8 ran the Oracle Solaris 11.1 operating system and used Oracle Solaris Zones to consolidate eight Oracle WebLogic application server instances and one database server instance to achieve this result. The IBM system used LPARS and AIX V7.1.This result demonstrated less than 1 second average response times for all SPECjEnterprise2010 transactions and represents JEE 5.0 transactions generated by 227,500 users.See benchmark disclosure slide
9 SPARC T5-2 SPECjbb2013 Multi-JVM World Record 2-chip on new realistic enterprise Java benchmarkSPECjbb2013 per-chipSPARC T5 1.9x faster HP x86SPARC T5 1.9x faster than HP E5-2690SPARC T5 2.23x faster than HP E7-4870IBM Power7+ only publishes SPECjbb2005, SPEC retiring old benchmark in July 2013SPARC T5Application ServerMulti-JVM resultsProcessorMax jOPSCrit jOPSHP DL980 G7 8-chip2.4GHz E7-4870106,14123,268SPARC T5-2 (2-chip)3.6GHz SPARC T575,65823,334HP DL560p 4-chip2.7GHz E5-465067,85016,199HP ML350p 2-chip2.9GHz E5-269040,04712,566SPARC T4-22.85GHz SPARC T434,80410,101• Oracle's SPARC T5-2 server achieved 75,658 SPECjbb2013-MultiJVM max-jOPS and 23,268 SPECjbb2013-MultiJVM critical-jOPS on the SPECjbb2013 benchmark. This result is a two chip world record. (Oracle has submitted this result for review by SPEC.)• There are no IBM Power7 or Power7+ based results on this important SPECjbb2013 benchmark. IBM has published IBM Power7+ based servers on the SPECjbb2005 which will soon be retired by SPEC.• The SPARC T5-2 server running Oracle Solaris 11 is 1.9x faster than the HP ML350p 2-chip (2.9 GHz E 'Sandy Bridge'-based server) based on SPECjbb2013-MultiJVM max-jOPS.• The 2-chip SPARC T5-2 server running Oracle Solaris 11 is 15% faster than the 4-chip HP ProLiant DL560p Gen8 (2.7 GHz E 'Sandy Bridge'-based server) based on SPECjbb2013-MultiJVM max-jOPS.• The 2-chip SPARC T5-2 server running Oracle Solaris 11 is 6.1 faster than the 1-chip HP ProLiant ML310e Gen8 (3.6 GHz E3-1280v2 'Ivy Bridge’) based on SPECjbb2013-MultiJVM max-jOPS.• Oracle's SPARC T4-2 server achieved 34,804 SPECjbb2013-MultiJVM max-jOPS and 10,101 SPECjbb2013-MultiJVM critical-jOPS on the SPECjbb2013 benchmark. (Oracle has submitted this result for review by SPEC.)• Oracle's Sun Server X3-2 system achieved 41,954 SPECjbb2013-MultiJVM max-jOPS and 13,305 SPECjbb2013-MultiJVM critical-jOPS on the SPECjbb2013 benchmark. (Oracle has submitted this result for review by SPEC.)• Oracle's Sun Server X2-4 system achieved 65,211 SPECjbb2013-MultiJVM max-jOPS and 22,057 SPECjbb2013-MultiJVM critical-jOPS on the SPECjbb2013 benchmark. (Oracle has submitted this result for review by SPEC.)• SPECjbb2013 demonstrates better performance on Oracle hardware and software, engineered to work together, than alternatives from HP.• Oracle's results were obtained using Oracle Solaris 11 along with Java Platform, Standard Edition, JDK 7 Update 17 on the SPARC T5-2 server and Java Platform, Standard Edition, JDK 7 Update 21 on the SPARC T4-2 server, Sun Server X3-2 and Sun Server X2-4.Xeon E5-2690Sandy BridgeSPARC T4Xeon E7WestmereSee benchmark disclosure slide
11 Typical Business Application Stack Challenges Current State Expensive to build and maintainComplex, multi-vendor architecture“One off” buildsHarder to scaleUnused capacitySiloed systemsHarder to troubleshootMore vulnerable
12 Reduced Complexity, Reduced Cost Hardware and Software Engineered to Work TogetherVS.For a long time Oracle’s vision included the notion of taking all the pieces of a system, all the hardware, and all of the software pieces, and before they’re designed, give some thought about how they would work together. As we mentioned earlier, Oracle has spent many years designing and building the industry’s most complete and consistent architecture for our enterprise software and systems products. That architecture is called the Oracle Fusion Architecture. Oracle Fusion Architecture is a standards-based technology blueprint that details the linkage between all Oracle products.Oracle is the only company that develops storage products, designs microprocessors and server platforms, as well as database and a comprehensive portfolio of business applications. When all those disparate pieces are built by design teams working toward the same vision, the result is an integrated system that dramatically improves performance, reduces costs and downtime, and provides customers with a more easily managed infrastructure.121212
13 Oracle Engineered Systems’ approach Designed as Complete InfrastructureTypical Custom ConfigurationPerformance AchievementMonthsLive to ProductionAssemble dozens of componentsMeasure, diagnose, tune, and reconfigureMulti-vendor testingTest and debug production failure modesWeeksOracle Engineered SystemEngineered Systems enable users to quickly adopt new technology that delivers game changing performance achievement in previously unheard of time frames. Oracle can reduce deployment and performance achievement via an Engineered System in just weeks as opposed to many month for typical complex custom configurations. Oracle Engineered Systems eliminate many of the most time and resource intensive stages of solution deployment by delivering a single configuration that has already been tested, tuned, configured, and completely engineered to work together in an optimal manner. No need to assemble components, integration test, debug, etc. Oracle has already done all of this at the engineering level to deliver the best solution with the best results in the least amount of time.
14 Engineered Systems Expedited time to value ExadataDatabase MachineExalogicElastic CloudExalyticsBig Data ApplianceSPARCSuperClusterWe are excited to launch 4 new products that joined Exadata and Exalogic in the Engineered Systems family earlier this year. These new products adhere to the same technology principles as Exadata and Exalogic, and extend the Engineered Systems’ value proposition.We have been in this business delivering Engineered Systems since 2008 with the introduction of Exadata followed closely by Exalogic. The success of these products has been truly amazing – outpacing even our goals for adoption. We have over 1200 systems sold to date and expect that number to be 4000 later this year. Exalogic has already surpassed Exadata as the fastest growing product in Oracle’s history and the demand for our new products leads us to believe we will see even greater adoption of our products as they were designed to be “Better Together”.With all of our products the deployment time is minimized, manageability is maximized, the lowest total cost of ownership in comparison to our competitors is delivered to you, implementation risk is minimized, and as always you have one single point of contact for support. These core principles apply to our existing products, our new products, and will certainly apply to the innovation we continue to deliver.Expedited time to valueEasier to manage and upgradeLower cost of ownershipReduced change management riskBetter reliability and one-stop supportExtreme performance
15 SPARC SuperCluster Attributes Applications plus database in one systemIntegrated and optimized Exadata Storage Servers for database performanceTrue “Oracle Best on Oracle”Zero overhead, most efficient virtualizationIndustry-fastest processor for end-to-end security (fastest Database)
16 SPARC SuperCluster T4-4 Best for Oracle. Runs Your Existing Workloads SPARC T4 Compute Pool More than a 20 World Records over IBM and HP across every tierExadata Storage Servers 750k IOPS, 43 GB/s query throughputExalogic Elastic Cloud 10x Java performanceIntegrated ZFS Storage 2x faster and 2x better price performance of NetAppOracle Solaris Cloud provisioning in secondsUnmatched ScalabilityVirtualization Oracle VM Server for SPARCNo cost and near zero overheadInfiniBand 5-8x the speed of current networksEnterprise Manager Up to 90% reduction of downtime due to proactive critical application patchingExadata purpose built for the database runs that extraordinarily well, and Exalogic runs middleware, Java middleware, WebLogic and those applications extraordinarily well. But what we wanted to do here is to bring those pieces of technology together with any other stuff you have running in your enterprise.So, we have many customers that have hundreds or thousands of applications, whether they’re packaged applications or written in traditional languages, running many parts of their enterprise. And we bring together these core technologies into a platform that lets you run those applications, but also have the best of the best from Oracle so you can run side-by-side with database 11, you can run side-by-side with WebLogic, and so on.Performance substantiation:1M IOPS: Based on internal measurement of Exadata Storage Servers10x Java performance: Based on internal measurement of ExalogicZFS 2x faster, ½ the price of NetApp: Demonstrates the performance of ZFS Storage via the Oracle Sun ZFS Storage 7420 appliance which delivered outstanding performance and price/performance on the SPC Benchmark 1, beating results published on the NetApp FAS3270A.The Sun ZFS Storage 7420 appliance delivered 137, SPC-1 IOPS at $2.99 $/SPC-1 IOPS on the SPC-1 benchmark. The Sun ZFS Storage 7420 appliance outperformed the NetApp FAS3270A by 2x on the SPC-1 benchmark. The Sun ZFS Storage 7420 appliance outperformed the NetApp FAS3270A by 2.5x on price/performance on the SPC-1 benchmark.SPC-1, SPC-1 IOPS, $/SPC-1 IOPS reg tm of Storage Performance Council (SPC). More info Oracle Sun ZFS Storage 7420 Appliance As of October 3, 2011.Cloud provisioning, unmatched scalability: Scalability of Solaris 11, leader in scaling to 512 threadsNear zero virtualization overhead: Solaris Zones, based on internal testsInfiniBand: 5-8x speed of current networksWhy Oracle Software Runs Best on Oracle Hardware
17 SPARC SuperCluster T5-8 Powered by Fastest Database and Applications Processor SuperCluster Gets Even BetterPowered by T5 with #1 performance for Database, Java, middleware, ERP, CRM etc~2.3x improvement in compute performance per rack33% more Exadata Storage ServersMaintains the same SSC architectureDramatic improvement in price performance
20 Exadata Storage Servers SPARC SuperCluster T5-8 Summary of ConfigurationsSSC T5-8 Half RackSSC T5-8 Full RackT5-8 Nodes2T5 Compute Cores128 (64 cores per node)256 (128 cores per node)T5 Memory2TB (1TB per node)4TB (2TB per node)Exadata Storage Servers48ZFS SA60TB~2.3x Compute capacity and 33% more Exadata Storage for both Full Rack and Half Rack SSC T5-8 vs SSC T4-4
22 Oracle Corporation Oracle E-Business Suite: Global Single Instance Database Platform SPARC SUPERCLUSTER BENEFITSRapid Deployment: 10 weeks from test to go liveTypical savings of $1.6M in hardware, software, and support costs over 5 yearsPERFORMANCE RESULTS4x average improvement in EBS batch times7.7x improvement in number of database read/write I/Os due to Exadata flash cache4-6x improvement in average I/O response time3.3x improvement in database log sync timeCOMPANY OVERVIEWOracle’s ERP implementation of Enterprise Business Suite R12 in a global single instancePrevious Implementation was 3 x M9000s and 2 x EMC DMX3000-M2 duplicated for production, disaster recovery, and testChallenges:I/O was saturated on EMC StorageEMC support was costlySOLUTIONSPARC SuperCluster plus 2 x Exadata Expansion Racks for production, disaster recovery, and testTypical savings based on 5 year TCO scenario for a commercial customer assuming 60% software discount. Exadata expansion racks included in replacement hardware costs.
23 Consumer Product Manufacturer, Asia IBM P Series Replacement, In Production COMPANY OVERVIEWManufacturing company in China significantly expanding technology infrastructure over the next 5 yearsStrategic plan to provide a highly available and elastic resources scaling infrastructure consolidate their multiple financial business applicationsPurchased SuperCluster as a complete turn key engineered solution over IBM’s proposal which requiring component integration at a higher costSPARC SUPERCLUSTER BENEFITSSuperior platform engineered optimized for Oracle Database and SPARC, fully integrated, with Oracle as the single vendor for support, with rapid time to deploymentOracle Solaris 10 and 11 determined as the most advanced and reliable OS for mission critical business, highly virtualized for consolidation with zoning mechanism allowing automatic policy-based resources allocationSPARC SuperCluster was chosen as the optimal platform over IBM’s proposal, since Oracle provided a single vendor, single footprint, centrally managedSOLUTIONSSPARC SuperCluster half rack running Oracle Enterprise Database and Web LogicSun Blade 6000 with SPARC T4-1B and X6270 M2 blade server for the application server tierCOMPETITIONThe competitive systems considered were IBM p770 and the IBM BladeCenter PS700 systemsSPARC SuperCluster was purchased over IBM's offer due to unique architectureLong-term plan to move all applications from Websphere/DB2 to WebLogic/OracleSPARC SUPERCLUSTER KEY DIFFERENTIATORSIntegrated hardware and software, leveraging a clustered database system with highly available failover capabilitiesTurn-key, integrated solution, rather than the traditional systems componentsOracle Solaris zoning and SPARC VM technologies provide cost saving for resources consolidation and virtualizationZoning mechanism allows automatic policy-based resources allocationThis account considered IBM pSeries and BladeCenter systems , and selected and purchased SPARC SuperCluster.Brief from the account team for presentor background:IBM proposed What solution including hardware and software did IBM propose? IBM's solution includes: a) 2x fully configured P770 as DB servers running DB2 database. b) BladeCenter with 8x PS700 and 4x HS22 as application servers running Websphere applications. 3. What made the customer decide to purchase SPARC SuperCluster instead of the IBM solution? In late 2011, the company decided to acquire a server platform that consolidates their multiple financial business applications, namely Accounting System, Budgeting System, Auditing System, Asset Management System, Capital Control and Capital Management System. The customer also formulated a strategic planning to provide a highly available and elastic resources scaling infrastructure for their business needs. Customer tends to buy a solution from one single vendor instead of integrating components from different vendors. Here are the reasons why customer chose SSC: 1) SSC is central to implement customer's infrastructure strategy as it well integrates hardware, software and leverages a clustered database system with highly available failover capabilities. 2) T+M Servers Competitive Landscape sales tool helps us compete effectively towards our rivals. 3) Successful selling strategy by positioning SSC as a highly optimized and turn key integrated solution rather than the traditional systems components selling. 4) Solaris provides the high end mission critical OS for business needs. 5) Oracle Solaris zoning and SPARC VM technologies provide cost saving for resources consolidation and virtualization. Moreover, the zoning mechanism allows automatic policy-based resources allocation. 6) Building strong working relationship with Apps team while driving one Red Stack into the account. 7) The half-rack SSC is relatively cheaper than IBM's solution while providing the similar computing capabilities in terms of hardware configuration. 8) Customer has made a long-term plan to move all applications from Websphere/DB2 to WebLogic/Oracle. 4. What were the primary features and benefits of SPARC SuperCluster over IBM? 1) Oracle database is well integrated and optimized in SSC. 2) Solaris 10 & 11 is a more advanced and reliable OS for mission critical business. 3) Oracle Solaris zone and Oracle VM server for SPARC technologies provide cost saving for resources consolidation and virtualization. Moreover, the zoning mechanism allows automatic policy-based resources allocation.
25 BackgroundEvolution of SPARC Systems and Oracle Solaris technologies :SPARC T5/M5/M6 SystemsSPARC SuperclusterVM2.0, ZFS, Zones, OVM, Critical Thread support, etc.ZFSSA, Exadata Storage, Flash TechnologiesInfiniband (IB) based Interconnect technologies
26 Oracle Database Integrations With Solaris Largest SGA Size on planet: 64TB is supportedDB startup optimizations with Solaris 11.1 VM2.0 subsystem: Improved large page availability and helps Oracle DB availability in generalOSM support in DB for dynamic sizing of SGA and robust manageabilityCritical Threads feature in Database for Logwriter and LMSIntegrated Crypto Acceleration for DB level crypto operations using ISA level support on T4/T5Zones and Ldoms for Better Oracle DB consolidationImproved RAC scalability with much reduced interconnect latencyImproved Observability and Profiling techniques
27 Oracle Database Integrations With Solaris Solaris 11.1 VM2.0 User BenefitsMemory demand predictor: enable better coordination and improved scalability for large page creation, cage growth, availabilityDistributed kernel cage: reduce latency as well as memory backplane usageCapture mechanism: common for multiple operations such as DR, FMA, cage growth, large page creation, and enable memoryMemory power managementTile data structures: improve freelist scalabilitySubstantial improvement in debug support
28 Oracle Database High Availability On Solaris Improving DB Startup Time by faster SGA AllocationEnable large (64 TB) SGA support for Database on SolarisSolaris VM2.0 improves large page availability and helps DB availabilityOracle Solaris parallelizes DB SGA allocation using kernel threads (VMTASKS)Default maximum number of vmtasks to the number of cores in the system
29 Oracle Database High Availability On Solaris Improve DB Startup/Shutdown Time More in Next Major DB ReleaseEnable Deferred SGA allocation for Oracle DB next major release out of the box on Solaris, can expect ~3x reduction in startup (no-mount) time w.r.t. Oracle 11gR2Startup no-mount for 30 TB SGA on M5-32 using Oracle DB next major release in 120 secImproving spawning background, foregrounds and PQ slave processes time
30 Oracle Database Critical Thread Support Oracle Solaris Scheduler ConceptsScheduling ClassesTS, FX, RTDynamically modified using priocntl (requires privileges)MPO aware load balancingBalanced home lgroup assignment and lgroup aware schedulerCMT aware schedulingCritical Thread optimization (available in Solaris 11, Solaris 10U10)Dynamically provides exclusive access to shared hardware resourcesEnabled by running critical software processes/threads at FX 60Observabilityproc tools (plgrp, prstat)
31 Oracle Database Critical Thread Support Improve Response Time and ThroughputCritical Threads support is integrated and is shipping for onwards. Logwriter and LMS will run FX 60 by default.Reduction in CPU utilizations and involuntary context switchesWith LMS running FX 60, cluster wait time is reduced and RAC scalability is improved
32 Oracle Database Critical Thread Support FX scheduling for response time improvementUse FX for foregrounds (OLTP) and PQ Slaves (DSS)Typically, upto a 5% improvement with FX over TSApply with caution for systems running mixed workloads (OLTP, DSS, Batch, Others)
33 Crypto Acceleration for Oracle Database TDE Hardware Assisted Crypto AccelerationIntegrated Solutions for TDE Accelerations: Oracle DB and Oracle Solaris optimized to access hardware crypto unit.T4 crypto's implementation allows direct userland accessPerformance: Encryption using T4 crypto has less than10% overhead compared to cleartextSavings: Crypto acceleration, included at No additional cpu costManagement: Centralized key management for all encrypted Oracle DatabaseTDE features are enabled out of the box
34 Oracle Solaris Container for DB Consolidation ZonesConsolidating multiple database instances into separate containers on the same system enables competing applications, such as OLTP and DSS workloads to run with Solaris resource allocationPerformance: Zero OverheadTechniques to improve Instance CagingLeverage Solaris resource management framework (Oracle )Enable processor set aware lgroup memory allocationIt supports RDS V3 for SPARC SuperCluster and Exadata
35 OVM for SPARC for DB Consolidation LDOMs: SPARC Built In VirtualizationFull virtual machine that runs an independent operating system instance and contains virtualized CPU, memory, storage, consolePerformance: Zero overhead for cpu and memoryDirect H/W access with static direct IO, zero overhead on performance
36 Improve Observability for Oracle Database Improve I/O Performance Diagnosability in Next Generation DatabaseDtrace tightly integrated with Database Monitoring FrameworkNew v$kernel_io_outlier view added for reporting IO response time breakup on ASM and raw storageSQL> desc v$kernel_io_outlier Name Null? Type TIMESTAMP NUMBER IO_SIZE NUMBER IO_OFFSET NUMBER DEVICE_NAME VARCHAR2(513) PROCESS_NAME VARCHAR2(64) TOTAL_LATENCY NUMBER SETUP_LATENCY NUMBER QUEUE_TO_HBA_LATENCY NUMBER TRANSFER_LATENCY NUMBER CLEANUP_LATENCY NUMBER PID NUMBER CON_ID NUMBER
37 Improved Observability On Solaris Tools For Better Performance DiagnosabilitySystem performance : Solaris Platform Process/thread analysis : lockstat, plockstat, truss, prstat and pstackMemory Placement Optimization : plgrp, pmap, lgrpinfo Interrupt assignment and load balancing : mpstat, mdb, intrstatHardware utilization and capacity planning : cpustat, corestat, pgstat, pginfoAdvanced level of debugging : DtraceCorrelate data from different tools to identify systemic issuesFor e.g. high GC latencies <-> mpstat output <-> prstat outputCan lead us to scheduler issues or interrupt saturation, etc
38 Profiling Techniques DTrace Oracle Solaris Studio 12.3 Systemic analysis : End to end user to kernel to user Drill-down analysis : e.g. High syscalls, which system calls, who issues it , what are the callstacks, arguments to the calls etc. Speculative analysis Proc, Sched and IO providersOracle Solaris Studio 12.3Accelerate Application Performance : Compiler, Performance libraryImproves Observability : Performance Analyzer, DLight Enhances Developer Productivity : Developer Tools
40 Massive Solaris scalability for T5 and M5 Virtual MemoryThe T5 and M5 processors provide hardware acceleration for “TLB shootdowns”, thus demap operations in dynamic memory workloads is small on T5-8 and M5-32The VM2 project re-implemented the physical page management layer in Solaris 11.1, and offers scalability benefitsReplaced the VM system's global counters with credit trees in S11.1, and achieved a 45X speedup on an mmap() microbenchmark on T4-4 with 256 CPUsAdjust kernel’s VA to support large heap and more physical memory and adjust process VA's to support larger shared memory segmentsReduced contention on various locks by increasing lock array sizes and improving the object-to-lock hash functions.
41 Massive Solaris scalability for T5 and M5 Resource Limits As of Solaris 11.1, the default for “maxusers, max_nprocs, and pidmax” automatically scales up with CPU count and memory size, to a maximum of processes. As of Solaris 11.1, the default number of threads that can be created scales with CPU and memory to a maximum of 1.6MScheduler: All functionality associated with clock scales nicelyIn Solaris 11, eliminated the need to acquire the process and thread locks in most cases during quanta expiration and accounting,Eliminated or reduced the impact of several smallish O(N) calculations that had become significant at 1536 CPUs.
42 Massive Solaris scalability for T5 and M5 Device: Higher device throughput and/or lower CPU utilization per unit of throughput on larger systems.Re-implemented DVMA allocation using the Solaris kmem_cache and vmem facilities, with object size and quanta chosen to match common DMA transfer sizes. This provides a good balance between contention-free per-CPU caching, and redistribution of free space in the back end magazine and slab layers.Also modified drivers to use DMA pools more efficiently,Modified the IOMMU code so that 2GB of VA is available per PCI function, rather than per PCI root port.
43 Conclusions: Solaris and Oracle SPARC Oracle #1 in real-world performanceComplete datacenter – Oracle runs Oracle bestSPARC T5/M5 performs best on complex workloads because of designSeeOracle Solaris: unmatched reliability, security & massive scalability on very large systems
44 Required Benchmark Disclosure Statement Must be in SPARC T5 & M5 Presos with Benchmark Results (1 of 2)Copyright 2013, Oracle &/or its affiliates. All rights reserved. Oracle & Java are registered trademarks of Oracle &/or its affiliates.Other names may be trademarks of their respective owners..TPC Benchmark C, tpmC, and TPC-C are trademarks of the Transaction Processing Performance Council (TPC). SPARC T5-8 (8/128/1024) with Oracle Database 11g Release 2 Enterprise Edition with Partitioning, 8,552,523 tpmC, $0.55 USD/tpmC, available 9/25/2013, New Order 90th% Response Time 0.410sec. IBM Power 780 Cluster (24/192/768) with DB2 ESE 9.7, 10,366,254 tpmC, $1.38 USD/tpmC, available 10/13/2010, New Order 90th% Response Time 2.10 sec. IBM x3850 X5 (4/40/80) with DB2 ESE 9.7, 3,014,684 tpmC, $0.59 USD/tpmC, available 7/11/2011. IBM x3850 X5 (4/32/64) with DB2 ESE 9.7, 2,308,099 tpmC, $0.60 USD/tpmC, available 5/20/2011. IBM Flex x240 (2/16/32) with DB2 ESE 9.7, 1,503,544 tpmC, $0.53 USD/tpmC, available 8/16/2012. IBM Power 780 (2/8/32) with IBM DB2 9.5, 1,200,011 tpmC, $0.69 USD/tpmC, available 10/13/2010. Source: results as of 3/26/2013.SPEC and the benchmark name SPECjEnterprise are registered trademarks of the Standard Performance Evaluation Corporation. Results from as of 3/26/2013. SPARC T5-8, 57, SPECjEnterprise2010 EjOPS; SPARC T4-4, 40, SPECjEnterprise2010 EjOPS; Sun Server X2-8, 27, SPECjEnterprise2010 EjOPS; Cisco UCS B440 M2, 26, SPECjEnterprise2010 EjOPS; IBM Power 780, 16, SPECjEnterprise2010 EjOPS. IBM PowerLinux 7R2, 13, SPECjEnterprise2010 EjOPS. SPARC T SPECjEnterprise2010 EjOPS. SPARC T5-8 (SPARC T5-8 Server base package, 8xSPARC T5 16-core processors, 128x16GB-1066 DIMMS, 2x600GB 10K RPM 2.5. SAS-2 HDD, 4x Power Cables) List Price $268,742. IBM Power 780 (IBM Power 780:9179 Model MHB, 8x3.86GHz 16-core, 64x one processor activation, 4xCEC Enclosure with IBM Bezel, I/O Backplane and System Midplane,16x 0/32GB DDR3 Memory (4x8GB) DIMMS-1066MHz Power7 CoD Memory, 12x Activation of 1 GB DDR3 Power7 Memory, 5x Activation of 100GB DDR3 Power7 Memory, 1x Disk/Media Backplane. 2x 146.8GB SAS 15K RPM 2.5. HDD (AIX/Linux only), 4x AC Power Supply 1725W) List Price $992,023. Source: Oracle.com and IBM.com, collected 03/18/2013. SPEC and the benchmark name SPECjEnterprise are registered trademarks of the Standard Performance Evaluation Corporation. Results from as of 5/1/2013. SPARC T5-8, 27, SPECjEnterprise2010 EjOPS; IBM Power 780, 10, SPECjEnterprise2010 EjOPS. Oracle server only hardware list price is $298,494 and total hardware plus software list price is $1,565,092 as of 4/24/2013. IBM server only HW list price is $835,555 and HW+SW cost of $2,174, and BM PowerLinux 7R2 server total hardware plus software cost of $819, based on public pricing from as of 4/24/2013.SPEC & the benchmark name SPECjbb are registered trademarks of Standard Performance Evaluation Corporation (SPEC). Results as of 3/26/2013, see for more information. SPARC T5-2 75,658 SPECjbb2013-MultiJVM max-jOPS, 23,334 SPECjbb2013-MultiJVM critical-jOPS. Sun Server X2-4 65,211 SPECjbb2013-MultiJVM max-jOPS, 22,057 SPECjbb2013-MultiJVM critical-jOPS. Sun Server X3-2 41,954 SPECjbb2013-MultiJVM max-jOPS, 13,305 SPECjbb2013-MultiJVM critical-jOPS. SPARC T4-2 34,804 SPECjbb2013-MultiJVM max-jOPS, 10,101 SPECjbb2013-MultiJVM critical-jOPS. HP ProLiant DL560p Gen8 66,007 SPECjbb2013-MultiJVM max-jOPS, 16,577 SPECjbb2013-MultiJVM critical-jOPS. HP ProLiant ML350p Gen8 40,047 SPECjbb2013-MultiJVM max-jOPS, 12,308 SPECjbb2013-MultiJVM critical-jOPS. Supermicro X8DTN+ 20,977 SPECjbb2013-MultiJVM max-jOPS, 6,188 SPECjbb2013-MultiJVM critical-jOPS. HP ProLiant ML310e Gen8 12,315 SPECjbb2013-MultiJVM max-jOPS, 2,908 SPECjbb2013-MultiJVM critical-jOPS. Intel R1304BT 6,198 SPECjbb2013-MultiJVM max-jOPS, 1,722 SPECjbb2013-MultiJVM critical-jOPS, HP DL980 G7 106,141 SPECjbb2013-MultiJVM max-jOPS, SPECjbb2013-MultiJVM critical-jOPS.This is the FIRST slide must be shown with each presentation that show benchmark results!
45 Required Benchmark Disclosure Statement Must be in SPARC T5 & M5 Presos with Benchmark Results (2 of 2)Copyright 2013, Oracle &/or its affiliates. All rights reserved. Oracle & Java are registered trademarks of Oracle &/or its affiliates.Other names may be trademarks of their respective owners..Two-tier SAP Sales and Distribution (SD) Standard Application benchmarks SAP Enhancement package 5 for SAP ERP 6.0 as of 3/26/13:SPARC M5-32 (32 processors, 192 cores, 1536 threads) 85,050 SAP SD users, 32 x 3.6 GHz SPARC M5, 4 TB memory, Oracle Database 11g, Oracle Solaris 11, Cert# SPARC T5-8 (8 processors, 128 cores, 1024 threads) 40,000 SAP SD users, 8 x 3.6 GHz SPARC T5, 2 TB memory, Oracle Database 11g, Oracle Solaris 11, Cert# IBM Power 760 (8 processors, 48 cores, 192 threads) 25,488 SAP SD users, 8 x 3.41 GHz IBM POWER7+, 1024 GB memory, DB2 10, AIX 7.1, Cert# Two-tier SAP Sales and Distribution (SD) Standard Application benchmarks SAP Enhancement package 4 for SAP ERP 6.0 as of 4/30/12:IBM Power 795 (32 processors, 256 cores, 1024 threads) 126,063 SAP SD users, 32 x 4 GHz IBM POWER7, 4 TB memory, DB2 9.7, AIX7.1, Cert# SPARC Enterprise Server M9000 (64 processors, 256 cores, 512 threads) 32,000 SAP SD users, 64 x 2.88 GHz SPARC64 VII, 1152 GB memory, Oracle Database 10g, Oracle Solaris 10, Cert# SAP, R/3, reg TM of SAP AG in Germany and other countries. More infoSPEC & benchmark names SPECfp, SPECint are registered trademarks of the Standard Performance Evaluation Corporation. Results as of March 26, 2013 from and this report. SPARC T5-8: 3750 SPECint_rate2006, 3490 SPECint_rate_base2006, 3020 SPECfp_rate2006, 2770 SPECfp_rate_base2006; SPARC T5-1B: 467 SPECint_rate2006, 436 SPECint_rate_base2006, 369 SPECfp_rate2006, 350 SPECfp_rate_base2006. IBM Power chip 3.92GHz: 2640 SPECint_rate2006. IBM Power 710 Express 1-chip 3.556GHz: 289 SPECint_rate2006.TPC-H, QphH, $/QphH are trademarks of Transaction Processing Performance Council (TPC). For more information, see results as of 6/7/13. SPARC T , available 9/24/13, 4 processors, 64 cores, 512 threads; SPARC T , available 5/31/12, 4 processors, 32 cores, 256 threads; SPARC Enterprise M , available 9/22/11, 64 processors, 256 cores, 256 threads; SPARC Enterprise M , available 12/9/10, 32 processors, 128 cores, 128 threads; IBM Power , available 11/30/11, 8 processors, 32 cores, 128 threads; HP ProLiant DL980 G7 162, available 10/13/10, 8 processors, 64 cores, 128 threads. ; HP ProLiant DL980 G7 158,108 available 4/15/13, 8 processors, 80 cores, 160 threads.This is the SECOND slide must be shown with each presentation that show benchmark results!