Database In- Memory and Exadata: Do I sgll need Exadata? Mathew Steinberg Exadata and Database In- Memory Product Management IOUG BIWA Summit January 27-29, 2014 Redwood City, CA Oracle ConfidenGal Internal/Restricted/Highly Restricted
Agenda The Real Time Enterprise Exadata Overview and Update Database In- Memory Overview E- Business Suite In- Memory Cost Management Do I sgll need Exadata? Oracle ConfidenGal Internal/Restricted/Highly Restricted 3
DATA- DRIVEN AGI LE AGILE The Real Time Enterprise EFFICIENT Oracle ConfidenGal Internal/Restricted/Highly Restricted 4
Market Trend à Real- Time Enterprise The Trends The Challenge GlobalizaGon Growth Online Margin 24-7 Changes compeggve and Profitability Dynamics The Need Real- Time Enterprise DATA- DRIVEN AGI LE AGILE EFFICIENT Oracle ConfidenGal Internal/Restricted/Highly Restricted 5
DBIM and Exadata Deliver the Real Time Enterprise Data Mart OLTP AND Analy7cs OLTP Data Warehouse Daily Conven7onal Data Center Approach - Keep transacgonal and analygc systems separate. - Do not run extensive analygcs on OLTP systems. Real Time Enterprise Deliver results on current data in real Gme Oracle ConfidenGal Internal/Restricted/Highly Restricted 6
Exadata Overview and Update Oracle ConfidenGal Internal/Restricted/Highly Restricted 7
Oracle Exadata Database Machine Vision The best placorm for all Oracle Database workloads Warehousing è OLTP è ConsolidaGon è In- Memory Database Latest, most advanced hardware Fully scale- out servers and intelligent storage with unified InfiniBand connecgvity and PCI flash Unique sojware that maximizes the Oracle Database Database opgmized compute, storage, and networking sodware dramagcally improves performance and cost Standardized, op7mized, hardened end- to- end By the core Oracle RDBMS development team 8
Exadata X5-2 Hardware Overview Complete OpGmized Fully Redundant Scale- Out Scale- Out 2- Socket Database Servers Fastest Xeon chips, 18- core, 256 GB to 768 GB DRAM Unified Ultra- Fast InfiniBand Network 40 Gb InfiniBand internal connecgvity 10 Gb or 1 Gb Ethernet data center connecgvity Scale- Out 2- Socket Storage Servers 16 Xeon cores per server enables DB offload to storage Extreme Flash (EF) Storage è 12.8 TB Ultra- Fast PCI Flash Drives or High Capacity (HC) Storage è 6.4 TB Ultra- Fast PCI Flash Cards + 48 TB SAS disks 9
How Exadata Features Improve Performance Smart Flash Cache Uses flash for physical I/O Smart Flash Log Uses flash to improve log I/O latency + Write- Back Flash Cache Writes to flash to improve write latency Smart Scan Runs porgons of a database query in storage IO Resource Manager (IORM) PrioriGzes I/O bandwidth based on database or workload within a database
Exadata X5 Benefits Fastest for All Workloads Fastest OLTP new Extreme Flash 4.1 Million OLTP IOPS / rack Flash opgmized redo writes Fastest Data Warehousing 263 GB/sec analygc throughput / rack SQL offload to storage Fastest In- Memory Database Queries billions of rows per sec/core Fastest Communica7on Using InfiniBand SQL opgmized protocols Most Cost Effec7ve Tiers Memory, Flash, and Disk With best data compression Speed Allows Small Exadata to Replace Huge Servers + Arrays Fast Virtual Machines Unique end- to- end priorigzagon Elas7c Scale- Out Configura7on End- to- End Integrated Mgmt Standardized, Most Supportable Highest Availability Redundant Scale- Out Hardware Fastest Failure Recovery Server, storage, and network Best MAA implementa7on RAC, ASM, RMAN, Data Guard End- to- End Failure Tes7ng In- Memory Fault Tolerance 11
Exadata ElasGc Scale- Out ConfiguraGon Enables Incremental Expansion and New Workloads: In- Memory DB, All- Flash OLTP X5-2 Quarter Rack X5-2 Database Servers X5-2 Storage Servers Extreme Flash EF HC Max 22 servers or 38U of servers High Capacity Start with a Quarter Rack (physical config) 2 Database Servers 3 Storage Servers Add DB or Storage servers Mix EF and HC storage servers in same rack No need for half rack upgrade, or full rack upgrade Assembled with requested servers by Oracle Or add servers incrementally at customer site Can add X5-2 servers to older (v2 to x4) machines Standard Configura7ons also available Eighth Rack, Quarter Rack, Half Rack, Full Rack Eighth to Quarter Upgrade (Compute, Storage, Both) 12
Example Workload OpGmized Exadata ConfiguraGons DB In- Memory Machine Many DB Servers, High DRAM few Storage Servers Extreme Flash OLTP Machine All- Flash IOPs enables capacity based OLTP sizing Data Warehousing Machine Large Storage High Compute for Parallel SQL 576 DB Cores 13.3 TB RAM 32 TB Flash 240 TB Disk 288 DB Cores 2 TB RAM 102 TB Flash 288 DB Cores 224 Storage Cores 90 TB Flash 672 TB Disk 16 Database Servers + 5 High Capacity Storage Servers 8 Database Servers + 8 Extreme Flash Storage Servers 8 Database Servers + 14 High Capacity Storage Servers 13
Exadata Cost EffecGve Tiering of Memory, Flash, and Disk Unique, Transparent, Database Op7mized Storage Tiering and Compression 6 TB DRAM 89 TB PCI FLASH 672 TB DISK Hopest Data AcGve Data Cold Data Per standard 8 compute 14 storage HC DB Machine Full Rack Accelerates ALL Database workloads Exadata transparently delivers best of memory, flash and disk Speed of In- Memory DB I/Os of Scale- Out PCI Flash Cost of SAS Disk Storage Compression reduces cost and improves capacity Hybrid Columnar, In- Memory, OLTP Exadata speed and capacity allows it to replace huge servers and arrays I/O and InfiniBand speed allow higher CPU uglizagon reducing DB licenses 14
Exadata Dual- Format Columnar Flash Cache Hybrid Column (HCC) balances OLTP and analygcs 10X compression, fast analygc scans, single I/O OLTP As CPUs get faster, want even faster scans During Scans, Smart Flash Cache automagcally transforms HCC data to pure columnar as it is loaded into flash cache to further speed analygcs Only selected columns read from flash during a query Up to 5x query speedup Columnar cache also available on Extreme Flash Storage select columna from table where Flash Cache PopulaGon Completely automagc and transparent Requires Database 12.1.0.2 Compression Units Columns 15
Deploying ApplicaGons in the Cloud Exadata for Consolida7on and Database as a Service Best Mixed Workload Performance, No Bohlenecks, Performance Isola7on, Availability Any bopleneck on consolidated system can stall all workloads. Exadata eliminates boplenecks Highest network bandwidth, storage offload Millions of I/Os per second, unique log opgmizagons Exadata uniquely priorigzes I/O by pluggable database, job, user, service, etc. Exadata uniquely priorigzes crigcal DB network messages through engre fabric Exadata uniquely unifies CPU priorigzagon with I/O priorigzagon for end- to- end assurance Marke7ng Human Resources Manufacturing IT/ Opera7ons Engineering Sales Service Finance and Accoun7ng 16
Database In- Memory Oracle ConfidenGal Internal/Restricted/Highly Restricted 17
AnalyGcs is about finding paperns & trends by aggregagng data Oracle ConfidenGal Internal/Restricted/Highly Restricted 18
Row Format Databases vs. Column Format Databases Row SALES Transac7ons run faster on row format Example: Query or Insert a sales order Fast processing few rows, many columns Column SALES Analy7cs run faster on column format Example : Report on sales totals by region Fast accessing few columns, many rows 19
Row Format Databases vs. Column Format Databases Row Query SALES Query a single sales order in row format One contiguous row accessed = FAST Column SALES Stores Query a sales order in Column Format Many column accessed = S L O W Query Query Query Query Un7l Now Must Choose One Format and Suffer Tradeoffs 20
Oracle Database In- Memory Breakthrough: Dual Format Database Normal Buffer Cache SALES Row Format SALES New In- Memory Format SALES Column Format Transac7ons and Analy7cs BOTH row and column formats for same table Transac7ons run faster on row format Analy7cs run faster on column format Simultaneously acgve and transacgonally consistent AnalyGcs & reporgng use new in- memory Column format Data warehouses, data stores, data marts OLTP with analygc reporgng OLTP uses proven row format Oracle ConfidenGal Internal/Restricted/Highly Restricted 21
What is In- Memory REALLY good for? Speeds up analy7c data access What it doesn t help with Network round trips, logon/logoff Parsing, PL/SQL, complex funcgons Data processing (as opposed to access) Complex joins or aggregagons where not much data is filtered before processing Load and select once Staging tables, ETL, temp tables Etc. Understand the problem before recommending a solugon 22
What Makes an App a Good Candidate for In- Memory? Data is processed in the database not the applicagon OpGmizer picks correct plan Accurate stagsgcs exist Query accesses subset of columns Queries contain where clause predicates that greatly reduce the number of rows returned 23
Why Oracle Database In- Memory? #1 OLTP Scales with Unique Dual Format #2 Both Scale Up & Scale Out supported #3 Maximum Availability is Built In #4 Database Size Isn t Limited by Memory #5 All ApplicaGons are Transparently CompaGble #6 No new Personnel Needed
Now we can run Gme- sensigve analygcal queries directly against our OLTP database. This is something we wouldn't have dreamt of earlier. Arup Nanda, Enterprise Architect Starwood Hotels and Resorts 25
In- Memory on Exadata Oracle ConfidenGal Internal/Restricted/Highly Restricted 26
In- Memory Speed + Capacity of Low Cost Disk DRAM PCI FLASH DISK Hopest Data AcGve Data Cold Data Size not limited by memory Data transparently accessed across tiers Each tier has specialized algorithms & compression Simultaneously Achieve: Speed I/Os Cost of DRAM of Flash of Disk 27
Exadata Dual- Format Columnar Flash Cache Hybrid Column (HCC) balances OLTP and analygcs 10X compression, fast analygc scans, single I/O OLTP As CPUs get faster, want even faster scans During Scans, Smart Flash Cache automagcally transforms HCC data to pure columnar as it is loaded into flash cache to further speed analygcs Only selected columns read from flash during a query Up to 5x query speedup Columnar cache also available on Extreme Flash Storage select columna from table where Flash Cache PopulaGon Completely automagc and transparent Requires Database 12.1.0.2 Compression Units Columns 28
Scale- Out In- Memory Database to Any Size Scale-Out across servers to grow memory and CPUs In-Memory queries parallelized across servers to access local column data Direct-to-wire InfiniBand protocol speeds messaging on Engineered Systems 29
Oracle In- Memory: Industrial Strength Availability Data Guard & GoldenGate RAC ASM RMAN Pure In-Memory format does not change Oracle s storage format, logging, backup, recovery, etc. All Oracle s proven availability technologies work transparently Protection from all failures Node, site, corruption, human error, etc. 30
Oracle Database In- Memory: Unique Fault Tolerance Similar to storage mirroring Duplicate in-memory columns on another node Enabled per table/partition E.g. only recent data Application transparent Only Available on Engineered Systems Downtime eliminated by using duplicate after failure 31
E- Business Suite In- Memory Cost Management Cost Simulator Gross Profit Analyzer Cost Compare Oracle ConfidenGal Internal/Restricted/Highly Restricted 32
In- Memory Cost Management Cost and Profitability analysis in Real- 7me Supports Process and Discrete industries Provides real 7me analysis OpGmize costs and working capital Evaluate COGS and valuagons Maximizing margins and gross profits Spread sheet based UI Powered by Database In- Memory & Oracle Engineered Systems 12c Database In- Memory opgon Heap memory, Flash cache, write back flash cache Hybrid columnar storage and Data compression Process large volumes of cost data in real- Gme 33
In- Memory Cost Management Performance Benchmarks Key Benefits Significant performance gains with Oracle 12c Database In- Memory opgon Achieved 145x faster performance with 12c Database In- Memory opgon on commodity hardware Achieved over 1003x faster performance with Oracle 12c Database In- Memory opgon on Exadata X4-2 Time (minutes) 3470 3460 3450 3440 3430 50 40 30 20 10 From 58hrs to 3mins 27sec 58 hours (3460 mins) 23mins 56sec 3mins 27sec EBS Cost Rollup without 12C In- memory opgon With 12c In- memory opgon Commodity HW With 12c In- memory opgon Exadata X4-2 34
Do I SGll Need Exadata? Oracle ConfidenGal Internal/Restricted/Highly Restricted 35
Exadata and Database In- Memory Complementary Technologies Repor7ng Batch Exadata OLTP Data Warehouse Consolida7on DBaaS Database In- Memory Analy7cs Exadata accelerates mulgple (mixed) workloads, including analygcs Workloads generally are mixed Database In- Memory accelerates analygcs Together Exadata and Database In- Memory further accelerate analygcs OpGmized Storage Tiering and Compression Dual Format Flash Cache Exadata brings the following to Database In- Memory: Scalability In- Memory Fault Tolerance ElasGc configuragons opgmized for DBIM Oracle ConfidenGal Internal/Restricted/Highly Restricted 36
Oracle Database In- Memory on Exadata YES, you s7ll need Exadata Exadata accelerates mul7ple (mixed) workloads, including analygcs Exadata extends fast analy7cs to flash and disk for higher capacity Exadata is opgmal playorm for scale- out In- Memory clusters Fault Tolerance unique to Exadata Exadata provides high write I/O, and log I/O rates that are crigcal for databases that take advantage of In- Memory Exadata is configured, tuned, hardened, and tested, for DB workloads Oracle ConfidenGal Internal/Restricted/Highly Restricted 37
Oracle ConfidenGal Internal/Restricted/Highly Restricted 38