A Technical Overview of Oracle SuperClusterORACLE WHITE PAPER APRIL 2016

Table of ContentsIntroduction2Oracle SuperCluster Components4Oracle's SPARC M7 Component4Oracle's SPARC M6-32 Component4Common Oracle SuperCluster Components5Oracle SuperCluster Component Details6SPARC M7 Servers6SPARC M6-32 Servers8Oracle Exadata Storage Server11Oracle ZFS Storage Appliance12Networking—Oracle’s Sun Datacenter InfiniBand Switch 3613Oracle Solaris14Virtualization on Oracle SuperCluster15Secure Multitenancy on Oracle SuperCluster16Out-of-the-Box Security Controls17Oracle Enterprise Manager Ops Center17Oracle Solaris Cluster19Oracle's Exalogic Elastic Cloud Software (Optional)20Oracle SuperCluster M7 Configurations21Oracle SuperCluster M6-32 Configurations23Configuration Notes27Conclusion291 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

IntroductionOracle SuperCluster is Oracle’s fastest, most secure, and most scalable engineered system. It is acomplete engineered system for optimally running databases and applications on a single integratedsystem. Oracle SuperCluster is ideal for massive consolidation and private clouds. It is designed,tested, and integrated to run business-critical enterprise applications and rapidly deploy cloud serviceswhile delivering the highest levels of security, extreme efficiency, cost savings, and performance. Tomeet the security needs of the modern data center, Oracle SuperCluster employs a comprehensivedefense-in-depth security strategy that spans the database, application, server, storage, andnetworking components. As a result of its architectural design, Oracle SuperCluster models are wellsuited for multitier enterprise applications with web, database, and application components. Thisversatility, along with powerful, bundled virtualization capabilities, makes them ideal systems on whichto consolidate large numbers of applications, databases, and middleware workloads, or to deploycomplex, multiuser development, test, and deployment environments. They combine highly availableand scalable technologies, such as optional Oracle Database 12c and Oracle Database 11g withOracle Real Application Clusters (Oracle RAC) and optional Oracle Solaris Cluster software. Thearchitectures enable a high degree of isolation between concurrently deployed applications, which mayhave varied security, reliability, and performance requirements.Oracle SuperCluster provides optimal solutions for all database workloads, ranging from scanintensive data warehouse applications to highly concurrent online transaction processing (OLTP)applications. With their combination of Oracle Exadata Storage Server, Oracle Database software, andthe latest hardware components, Oracle SuperCluster M7 and Oracle SuperCluster M6-32 deliverextreme performance in a highly available, highly secure environment. Each Oracle Exadata StorageServer uses Exadata Smart Flash Logging, a feature of Oracle Exadata that both improves usertransaction response times and increases overall database throughput for I/O-intensive workloads byaccelerating performance-critical database algorithms.2 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

For example, it provides more processing power with up to 32 SPARC M6 processors from Oracle,more memory with up to 1 TB per physical SPARC M6 processor, more built-in I/O with 64 PCIe Gen 3slots, and, finally, three levels of virtualization: physical domains (or PDoms), logical domains (LDoms),and Oracle Solaris Zones.Customers can integrate Oracle SuperCluster systems with Oracle Exadata or Oracle Exalogicmachines by using the available InfiniBand expansion ports and optional data center switches. TheInfiniBand technology used by Oracle SuperCluster systems offer high-bandwidth, low latency,hardware-level reliability and security. For application environments that follow Oracle's best practicesfor highly scalable, fault-tolerant systems, no application architecture or design changes are requiredto benefit from Oracle SuperCluster systems. Deployments also can connect many OracleSuperCluster M7 and Oracle SuperCluster M6-32 systems or a combination of Oracle SuperClustersystems and Oracle Exadata to develop a single, large-scale environment sharing the same InfiniBandfabric. Customers can integrate Oracle SuperCluster systems with their current data centerinfrastructure using the available 10 GbE ports in each of Oracle’s SPARC servers within OracleSuperCluster.3 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

Oracle SuperCluster ComponentsOracle SuperCluster is comprised of many Oracle products that are preconfigured, pretuned, and pretested byOracle experts, eliminating weeks or months of effort typically required to design, integrate, and deploy a highperformance and highly available system. Extensive end-to-end testing ensures all components work seamlesslytogether and there are no performance bottlenecks or single points of failure that can affect the system. The maindifferences between Oracle SuperCluster M7 and Oracle SuperCluster M6-32 is the SPARC server.Oracle's SPARC M7 ComponentOracle SuperCluster M7 is a complete, preconfigured, cluster-capable solution utilizing the following components:» Oracle's SPARC M7 Servers—Oracle’s new SPARC M7 processor–based servers take Oracle’s servertechnology to new levels by offering the world’s first implementation of Oracle’s Software in Silicon technology tobuild clouds with the most secure platforms in the world. Offering both database and application security andacceleration, these servers offer at their core Software in Silicon features like Silicon Secured Memory and InMemory Query Acceleration, along with data compression and decompression and encryption.Each SPARC M7 chassis has the following:» Utilization of Oracle virtualization technology to electrically isolate the SPARC M7 chassis into twocompute nodes (physical domains)—Each compute node is equivalent to a physical server and can beconfigured with one, two, or four SPARC M7 processors.» Minimum of 512 GB and a maximum of 2 TB of memory per SPARC M7 compute node—512 GB (32GB dual inline memory modules [DIMMs]) of memory is provided per SPARC M7 processor.» Minimum of two compute nodes and maximum of four compute nodes per rack—Each SPARC M7chassis can contain one or two compute nodes, and a rack can contain up to two SPARC M7 chassis for amaximum of four compute nodes. A minimum of two compute nodes is required for application or databasehigh availability.» Oracle’s Sun PCIe dual-port QDR InfiniBand host channel adapter (HCA)—This is a low-latency 40Gb/sec InfiniBand HCA. There is one InfiniBand HCA configured per processor.» 10 Gigabit Ethernet adapters—One 10 GbE adapter is configured per processor.Oracle's SPARC M6-32 ComponentOracle SuperCluster M6-32 is a complete, preconfigured, cluster-capable solution utilizing the followingcomponents:» SPARC M6-32 servers—Oracle’s SPARC M6-32 server offers a large memory capacity and a highly integrateddesign that supports virtualization and consolidation of mission-critical applications. Oracle SuperCluster M6-32can be flexibly configured to meet the most demanding data center workloads.Each SPARC M6-32 server has the following:» Utilization of Oracle virtualization technology to electrically isolate the SPARC M6-32 rack into two orfour compute nodes (physical domains)—Each compute node can be configured with 4, 8, 12, or 16SPARC M6 processors based on specific compute node configurations.» Minimum of 4 TB and a maximum of 32 TB of memory per SPARC M6-32 rack—512 GB (16 GBDIMMs) or 1 TB (32 GB DIMMs) of memory is provided per SPARC M6 processor.» Minimum of 8 and a maximum of 32 SPARC M6 processors per SPARC M6-32 rack—Each processorcomes with 12 cores and 8 threads per core.4 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

» Sixteen to 32 disk drives per rack—There are eight 1.2 TB 10,000 RPM SAS disk drives per I/O unit(IOU).» Oracle’s Sun PCIe dual-port QDR InfiniBand host channel adapter—This is a low-latency 40 Gb/secInfiniBand HCA. There are four InfiniBand HCAs in each IOU.» Eight to 16 base I/O cards—These cards provide SAS controllers for the disks in an IOU. There are four ineach IOU, and each card also provides two 10 GbE ports. The 10 GbE ports are for client access to theOracle SuperCluster M6-32.Common Oracle SuperCluster ComponentsBoth Oracle SuperCluster M7 and Oracle SuperCluster M6-32 offer these elements:» Oracle Exadata Storage Server—These servers provide three critical technologies to improve databaseperformance: the smart scale-out storage and the Exadata Smart Flash Cache and Exadata Hybrid ColumnarCompression features of Oracle Exadata. There are a minimum of three Oracle Exadata Storage Server systemsin an Oracle SuperCluster M7 base configuration and a minimum of nine in the Oracle SuperCluster M6-32 baseconfiguration. There are two storage technology options for Oracle Exadata Storage Server. The first is usingflash-based storage and the second is using high-capacity disks storage. Oracle SuperCluster can scale storagecapacity by configuring additional Oracle Exadata Storage Server systems.» Oracle ZFS Storage ZS3-ES appliance—Providing 160 TB of raw disk capacity, this appliance uses the flashenabled technology of Hybrid Storage Pool, a feature of Oracle ZFS Storage Appliance, to improve applicationresponse times. Its performance scalability for file-based I/O and ease of management make it a good fit formanaging shared application data files within Oracle SuperCluster.» Oracle’s Sun Datacenter InfiniBand Switch 36—This switch provides a high-throughput, low-latency, andscalable fabric suitable for fabric consolidation of interprocess communication, network, and storage. InfiniBanddelivers up to 63 percent higher transactions per second for Oracle Real Application Clusters (Oracle RAC) overGbE networks.» Ethernet management switch—This switch provides network management connectivity to the managementports on all servers and switches used in Oracle SuperCluster.» Oracle Solaris operating system—Oracle Solaris 11 is a secure, integrated, and open platform engineered forlarge-scale enterprise cloud environments, allowing users to deploy enterprise mission-critical applications safelyand securely with no compromise.» Integrated virtualization—Enhanced security, increased utilization, and improved reliability are delivered throughOracle Solaris Zones and through Oracle VM Server for SPARC (previously known as Oracle's Sun LogicalDomains).» Oracle Enterprise Manager Ops Center—Oracle Enterprise Manager Ops Center delivers a convergedhardware management solution that integrates management across the infrastructure stack to help IT managersdeploy and manage Oracle SuperCluster more efficiently.» Oracle's Exalogic Elastic Cloud Software (optional)—Oracle Exalogic provides extreme performance for Javaapplications, Oracle Applications, and all other enterprise applications, and it reduces application implementationcosts and ongoing costs compared to traditional enterprise application platforms and private clouds assembledfrom separately sourced components. Oracle SuperCluster configurations contain a specific combination ofservers, storage, network, and software elements to ensure the highest quality for integration and testing duringproduction. The systems can expand the amount of storage.5 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

Oracle SuperCluster Component DetailsAll components selected and integrated into the Oracle SuperCluster platform provide unique and differentiatedfeatures. These features are integrated and tested together to provide a feature-rich solution for the mostdemanding data center requirements.SPARC M7 ServersThe SPARC M7 servers are designed for modern cloud infrastructures. They are ideal for database and commercialbusiness applications requiring operational efficiency, reliability, and scalability for large mission-critical computingenvironments.Oracle’s new SPARC M7 processor–based servers take Oracle’s server technology to new levels by offering theworld’s first implementation of Oracle’s Software in Silicon technology to build clouds with the most secure platformsin the world. Offering both database and application security and acceleration, these servers offer Silicon SecuredMemory, In-Memory Query Acceleration, data compression and decompression, and encryption at their core.With its new Software in Silicon capabilities coupled with an innovative cache and memory hierarchy, Oracle’sSPARC M7 processor delivers dramatically higher processing speed and revolutionary protection against malwareand software errors.The per-thread performance is improved with the entirely new on-chip L2 and L3 cache design and increasedprocessor frequency. The 64 MB L3 cache is partitioned and fully shared, and hot cache lines are migrated to theclosest partition to minimize latency and maximize performance. The architecture of the core clusters and partitionedcache is ideal for server virtualization and pluggable databases. System administration and performance tuning areeasier, because the design minimizes interaction between logical domains or between databases. The processorcan dynamically trade per-thread performance for throughput by running up to 256 threads, or it can run fewerhigher-performance threads by devoting more resources to each thread. This flexibility allows the system to balanceoverall throughput versus per-thread performance for optimal results.The SPARC M7 processor incorporates eight on-chip accelerators to offload in-memory database query processingand perform real-time data decompression, while crypto instruction accelerators are integrated directly into eachprocessor core. Together, the Software in Silicon features deliver significant performance advantages, including thefollowing:» Silicon Secured Memory provides real-time data integrity checking to guard against pointer-related softwareerrors and malware, replacing very costly software instrumentation with low-overhead hardware monitoring.Silicon Secured Memory enables applications to identify erroneous or unauthorized memory access, diagnose thecause, and take appropriate recovery actions.» In-Memory Query Acceleration provided by the accelerators delivers performance that is up to 10 times fastercompared to other processors.» The In-Line Decompression feature enables storing up to three times more data in the same memory footprint,without a performance penalty.» Accelerated cryptography helps eliminate the performance and cost barriers typically associated with securecomputing—which is increasingly essential for modern business operation. These accelerators enable high-speedencryption for more than a dozen industry-standard ciphers, eliminating the performance and cost barrierstypically associated with secure computing.Each of the eight in-silicon data analytics accelerators (DAXs) or Software in Silicon coprocessors included on theSPARC M7 die provides the following:6 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

» A data/message pipe for very fast local data movement to offload the CPU and provide secure remote clustermessaging» A query pipe for in-memory columnar acceleration that scans data vectors and applies predicatesTogether, the coprocessors provide very low-overhead interprocess communication and very fast atomic operations.For example, DAXs located on different processors can exchange messages and access remote memory locations,exchanging locks without CPU involvement. Utilizing this functionality requires Oracle Database 12c with theOracle Database In-Memory option and Oracle Solaris 11.3 or later.CPU, Memory, and I/O Unit ChassisThe SPARC M7 server is contained within one CPU, memory, and I/O unit (CMIOU) chassis—the enclosure thathouses the processor/memory boards, service processors (SPs), and connectors for the interconnect assemblies.The enclosure also includes power supplies and cooling fans. Figure 1 shows the front and rear view of the CMIOUchassis and its key components.The SPARC M7 chassis used with Oracle SuperCluster M7 supports up to total of eight processors. However, theOracle SuperCluster M7 chassis consists of two electrically isolated hardware partitions. This is done using Oracle'sPhysical Domains, to create two isolated compute nodes within each chassis. Therefore the maximum number ofprocessors per compute node is four, and there are two compute nodes within a single chassis. All SPARC M7server hardware required to operate both compute nodes is contained in a single CMIOU chassis.The front of the chassis features hot-swappable fan modules and power supply units (PSUs) as well as interconnectassemblies that connect the CMIOU boards together into a system. Front-facing features include the following:» Eight hot-swappable fan modules» Six hot-swappable N N redundant PSUs, 3,000 W at 200 VAC to 240 VACThe rear of the CMIOU chassis includes the following:» Up to eight CMIOU boards» Two redundant SPsFigure 1. SPARC M7 chassis7 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

CPU, Memory, and I/O Unit BoardEach CMIOU board assembly contains one SPARC M7 processor on a mezzanine board plus associated memoryand I/O. All 16 memory DIMM slots are on the board. An I/O controller ASIC provides dedicated root complexes forthree PCIe 3.0 (x16) slots. PCIe hot-pluggable carriers are included with the board. When inserted into the CMIOUchassis, the CMIOU board connects with the interconnect assemblies that provide the connectivity between theCMIOU boards and SPs.The Oracle SuperCluster M7 chassis uses CMIOU boards with the following configuration:» One SPARC M7 processor with 32 cores @ 4.13 GHz» 512 GB of memory (16 x 32 GB DDR4 DIMMs)» One InfiniBand host channel adapter» One 10 GbE adapter» One four-port Ethernet adapter (only for the first CMIOU per compute node [physical domain])Interconnect AssemblyThe systems interconnect for the SPARC M7 server is implemented with the interconnect assemblies. The SPARCM7 server uses interconnect assemblies that are contained within the 10U space of the CMIOU chassis. Fiveinterconnect assemblies provide the coherency link (CL) connections (glueless systems interconnect) between theCMIOU boards. One interconnect assembly is used for communication between the CMIOU boards and the dualSPs at the top of the CMIOU chassis.Service Processor, Service Processor Proxy, and Service Processor ModuleThe SPARC M7 server features redundant hot-pluggable service processors (SPs). There are two SPs located inthe CMIOU chassis. Each SP has one serial (RJ45) and one 1000BASE-T management port (RJ45, autonegotiatesto 10/100/1000 Mb/sec). The SP communicates with the CMIOU boards via the SP interconnect assembly.Each SP includes two service processor modules (SPMs) in order to always have redundant SP functionality andfailover capability. The SPM is the component that runs the Oracle Integrated Lights Out Manager (Oracle ILOM)software and provides the SP functionality for the server system.SPARC M6-32 ServersThe SPARC M6-32 server provides high throughput and computing density along with built-in virtualization andextreme scalability and is a highly efficient platform for deploying large-scale, mission-critical applications.8 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

Figure 2. SPARC M6-32 serverArchitected to reduce planned and unplanned downtime, the SPARC M6-32 server includes advanced reliability,availability, and serviceability (RAS) capabilities to avoid outages and reduce recovery time. These design elementsare vital for a mission-critical system such as Oracle SuperCluster. Design features that boost the reliability of theSPARC M6-32 server include:» Advanced CPU integration—The SPARC M6 processor is a 12-core processor, with each core featuring eightthreads. Each processor provides 48 MB of L3 cache, which is critical in reducing the time required to read andprocess data.» Extended-ECC memory—The memory system has redundant components built on to each memory DIMM,which allow a DIMM to continue to operate even with partial failure. Utilizing 16 GB DIMMs or 32 GB DIMMs,each SPARC M6-32 server in Oracle SuperCluster M6-32 provides up to 16 TB or 32 TB of memory.» Memory lane sparing—Each of the two built-in memory controllers in the SPARC M6 processor has memorylane sparing. In the event that a particular lane fails, the memory controller can continue without interruption toaccess the main memory.» Processor lane sparing—The interconnect on all SPARC M6 processor/memory boards has lane sparing. Thismeans there are no lost packets of data during cache coherency or for remote memory access.» Fault-resilient power options and hot-swappable components—Systems feature redundant, hot-swappableservice processors, power supply units, and fan units. Redundant storage can be created using hot-swappabledisk drives with disk-mirroring software.» Hardware redundancy— SPARC M6-32 servers feature redundant system interconnect boards, redundant clockboards, redundant power, redundant fans, and redundant data paths.The 10 GbE interfaces on the base I/O cards are used by clients for access to the Oracle SuperCluster M6-32compute nodes. In addition to the 10 GbE interfaces, also included are Quad Data Rate (QDR) InfiniBand cards.These are used for internal communication and access to the instances of Oracle Exadata Storage Server and theOracle ZFS Storage ZS3-ES appliance.All CPUs and memory are contained on boards called CPU/memory units (CMUs). All I/O components are placed inan I/O unit (IOU). There is one IOU for every four CMUs. A grouping of four CMUs and an IOU is called a domain9 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

configuration unit (DCU). In the SPARC M6-32 server, four DCUs are possible, and each DCU can have either twoor four CMUs.Figure 3. DCU assignmentsThe DCUs are building blocks that are used to configure the physical domains (PDoms). In Oracle SuperCluster M632, there are either two or four PDoms, with two-PDom configurations supporting either one or two DCUs each.There are two types of configurations, basic and extended. Base configurations always have one DCU per PDom,and extended configurations always have two DCUs per PDom.Because each DCU has its own IOU, there are enough PCIe slots to provide I/O connectivity for the standard I/Ocards included. Each IOU has four QDR InfiniBand cards, four base I/O cards (also called Express Module SAS[EMS] cards) with two 10 GbE ports each, eight 1.2 TB 10,000 RPM SAS disk drives, and one 1 GbE card fordomain management. Figure 4 illustrates the IOU layout.Figure 4. IOU layout10 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

The 10 GbE ports on the EMS cards are the primary network ports for client access to Oracle SuperCluster M6-32.Platform management is done via the service processor, or Oracle Integrated Lights Out Manager (Oracle ILOM)3.0. Oracle ILOM provides a command-line interface (CLI), a web-based graphical user interface (GUI), andIntelligent Platform Management Interface (IPMI) functionality to aid out-of-band monitoring and administration. Themanagement software, Oracle Enterprise Manager Ops Center, communicates with Oracle ILOM to manage andmonitor Oracle SuperCluster M6-32.Oracle Exadata Storage ServerOracle Exadata Storage Server runs Oracle’s Exadata Storage Server Software, which provides unique andpowerful software technology, including its Exadata I/O Resource Manager and Exadata Storage Indexes features,as well as these Oracle Exadata features: Exadata Smart Scan, Exadata Smart Flash Cache, Exadata Smart FlashLogging, and Exadata Hybrid Columnar Compression. The hardware components of each Oracle Exadata StorageServer (also referred to as an Oracle Exadata cell) are carefully chosen to match the needs of high-performancedatabase processing. Exadata Storage Server Software is optimized to take the best possible advantage of thehardware components and Oracle Database. Each Oracle Exadata cell delivers outstanding I/O performance andbandwidth to the database. When used in Oracle SuperCluster, Oracle Exadata Storage Server can be accessedonly by Oracle Database 12c and Oracle Database 11g Release 2 running in the database domain. They cannot beused for any other purpose beyond Oracle Database 12c and Oracle Database 11g Release 2.Figure 5. Oracle Exadata Extreme Flash and High Capacity Storage ServerOracle Exadata Storage Server delivers exceptional database performance by way of the following criticaltechnologies:» Exadata Smart Scan―processes queries at the storage layer, returning only relevant rows and columns to thedatabase compute nodes. As a result, much less data travels over fast 40 Gb InfiniBand interconnects—dramatically improving the performance and concurrency of simple and complex queries.» Exadata Smart Flash Cache―addresses the disk random I/O bottleneck problem by transparently caching “hot”frequently accessed data to fast solid-state storage. It provides up to a 30x improvement in response time forreads compared to regular disk and up to 20x more write performance―a hundred-fold improvement in IOPS forreads compared to regular disk. And, it is a less expensive higher-capacity alternative to memory.» Exadata Smart Flash Logging―takes advantage of the flash memory in Oracle Exadata storage to speed up logwrites.» Exadata Hybrid Columnar Compression―can reduce the size of data warehousing tables by an average of 10xand of archive tables by 50x. This offers significant savings on disk space for primary, standby, and backupdatabases, and dramatically improves the performance of data warehousing queries.There are two Oracle Exadata Storage Server configuration options. The first is a high-capacity option with bothflash storage and 12 high-capacity SAS disks. The second is an extreme flash option using eight NVMe PCI flashdrives, providing maximum performance for the most demanding workloads.While Oracle Exadata Storage Server can be accessed only by a production Oracle Database 12c or OracleDatabase 11g Release 2 database running in a database domain, an Oracle Database 12c or Oracle Database 11g11 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

Release 2 database can be run in an application domain on Oracle Solaris 11 for testing and developmentpurposes. The Oracle ZFS Storage ZS3-ES appliance then can store the database. Exadata Hybrid ColumnarCompression software technology also can be used on the Oracle ZFS Storage ZS3-ES appliance to test ExadataHybrid Columnar Compression functionality before moving the database to the database domain and using OracleExadata Storage Server.Oracle ZFS Storage ApplianceFor shared file storage, Oracle SuperCluster includes an appliance from the Oracle ZFS Storage Appliance family,specifically the Oracle ZFS Storage ZS3-ES appliance, which features a common, easy-to-use managementinterface and the industry’s most comprehensive analytics environment. To deliver high performance using costeffective components, the Oracle ZFS Storage Appliance file system, Oracle Solaris ZFS, seamlessly optimizesaccess to the different types of media in the Hybrid Storage Pool technology. Oracle Solaris ZFS is designed toautomatically recognize different I/O patterns and place data in the best storage media for optimal performance.Oracle ZFS Storage Appliance in Oracle SuperCluster features a two-node cluster configuration that enables highperformance and high availability (HA) to maximize business productivity. This is the configuration used in OracleSuperCluster.Figure 6. Oracle ZFS Storage ZS3-ES appliance12 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

Figure 7. Real-time analytics help administrators increase storage optimization and reduce down time.For Oracle SuperCluster M7, Oracle ZFS Storage ZS3-ES appliance is installed in the main Oracle SuperCluster M7rack. For Oracle SuperCluster M6-32, Oracle ZFS Storage Appliance is installed in the Oracle SuperCluster M6-32storage rack. Oracle ZFS Storage ZS3-ES appliance can be made accessible from the database domain and usedfor Oracle Recovery Manager (Oracle RMAN) backups or flat file staging. Oracle RMAN is a feature of OracleDatabase.Networking—Oracle’s Sun Datacenter InfiniBand Switch 36InfiniBand technology has emerged as an attractive fabric for building large supercomputing grids and clusters. Asan open standard, InfiniBand presents a compelling choice compared to proprietary interconnect technologies thatdepend on the success and innovation of a single vendor.Figure 8. Sun Datacenter InfiniBand Switch 3613 A TECHNICAL OVERVIEW OF ORACLE SUPERCLUSTER

Oracle SuperCluster M7 contains either two or three Sun Datacenter InfiniBand Switch 36 switches. OracleSuperCluster M6-32 contains three Sun Datacenter InfiniBand Switch 36 switches. The first two switches are alwaysleaf switches and the third is used as a spine switch to connect multiple SuperCluster and other engineered systemracks together. The two leaf switches are connected to each other to provide redundancy in case one of the two leafswitches fail. In addition, each SPARC compute node, Oracle Exadata Storage Server, and each controller on theOracle ZFS Storage ZS3-ES appliance has connections to both leaf switches to provide redundancy in theInfiniBand connections in case one of the two leaf switches fail.Oracle SolarisOracle Solaris provides key functionality for virtualization, optimal use, high availability, unparalleled security, andextreme performance for both vertically and horizontally scaled environments. Oracle SuperCluster supports bothOracle Solaris 10 and Oracle Solaris 11 on Oracle SuperCluster M6-32. On the Oracle SuperCluster M7, OracleSolaris 11 is required for all domains while Oracle Solaris 10 can be used in a zone. Database domains exclusivelyuse Oracle Solaris 11 on both SuperCluster platforms.Oracle Solaris includes the following featur

Both Oracle SuperCluster M7 and Oracle SuperCluster M6-32 offer these elements: » Oracle Exadata Storage Server—These servers provide three critical technologies to improve database performance: the smart scale-out storage and the Exadata Smart Flash Cache and Exadata Hybrid Columnar Compression features of Oracle Exadata.