Oracle RAC - Customer Proven Scalability
1 like2,188 views
The document discusses Oracle Real Application Clusters (RAC) and provides examples of how it has enabled scalability and high availability for many large customers. It describes how RAC allows databases to scale horizontally across multiple servers, provides several customer cases that have implemented RAC with 4+ nodes, and highlights how RAC provides scalability by design through its instance and global cache architecture.
Oracle RAC - Customer Proven Scalability
- 2. Oracle Real Application Clusters (RAC) Customer Proven Scalability Copyright © 2014, Oracle and/or its affiliates. All rights reserved. Markus Michalewicz, Troy Anthony Oracle Real ApplicaEon Clusters (RAC) and DBaaS Product Management November 6th, 2014 @OracleRACpm h:p://www.linkedin.com/in/markusmichalewicz h:p://www.slideshare.net/MarkusMichalewicz
- 3. Oracle RAC Customer Examples Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 3
- 4. Oracle RAC Customers with 4+ Nodes 12 out of these 56 customers (~21%) use(d) 4 or more nodes Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 4
- 5. Copyright © 2014, Oracle and/or its affiliates. All rights reserved. • Does the applicaEon scale on a single (SMP) system? 5 Scaling Principles write write write write write write write write write write write write write write write write • If it does, it is likely to scale horizontally. • Scalability is measured considering the whole system.
- 6. Stock Exchange Workload Test 67%↑ Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 6 1 4 8 8 6 2 # of CPUs # of Nodes 6 640 tps 2 4 1200 tps 1200 tps 88%↑ 88%↑ 2200 tps 83%↑ 83%↑ 2208 tps 2000 tps 67%↑ 84%↑ 3600 tps 64%↑ 76%↑ 80%↑ 78%↑ 3888 tps 80%↑ 6480 tps 70%↑ 6424 tps
- 7. Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 7 Mercado Libre Scaled beyond 10 nodes already in 2007 2,000,000 1,500,000 1,000,000 500,000 0 Business Volume 2004 2005 2006 Nodes 2007
- 8. Oracle RAC = Scalability by Design Instance A Instance B Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 8 1 Instance C 2 GCS Current 225 Update Block 10 Requester Holder 200 3 Coordinator
- 9. Oracle RAC Customers – Strong on RAC Now and Then Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 9 In 2008 Yahoo standardized on Oracle RAC with one RAC database spanning at least 2, but no more than 8 nodes, which was also the maximum size of the cluster. “Full support for RAC scale-‐out means Oracle Database In-‐Memory can be used on our largest Data Warehouse, enabling more near real-‐Eme analyEcs.” • Sudhi Vijayakumar, Senior Oracle DBA Yahoo Inc. In 2014, Yahoo is sEll strong on Oracle RAC, adopEng new technologies.
- 10. Oracle RAC Customers – Strong on RAC Now and Then In 2014, Postbank is sEll strong on Oracle RAC, using commodity hardware and 10G. In 2007, Postbank tested with 4 nodes using Infiniband to determine the new standard setup for their data centers. Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 10
- 11. Burlington Coat Factory: RAC / ADG Benefits Be:er Performance 3 4 Copyright © 2014, Oracle and/or its affiliates. All rights reserved. “We have two data centers. They're ten kilometers apart. We have two clusters, and on each cluster we run the Oracle Data Guard copies for the other cluster. So basically all the data is always in both data centers in real Ame. ” -‐ Mike Price, CTO Business Objec[ves • High Server Load/ Poor System Availability. • Many independent two-‐node clusters impacEng IT staff ProducEvity • No Disaster Preparedness • Awkward Storage Provisioning and Management ResulEng in Outages, Uncertainty Around Remaining Capacity Solu[on • RAC on Suse Linux 50% faster queries Availanility Inc up[me and DR Scalable hardware Two 18 node clusters IT Produc[vity Simplified management using EM Data Center 1 Data Center 2 1 SKU DB 2 SKU DB SKU DB SKU DB Standby for SKU 5-‐8 DBs 5 SKU DB 7 SKU DB 6 SKU DB 8 SKU DB 10km 18-‐node cluster RAC IBM xSeries SUSE Linux Data Center 1 Data Center 2
- 12. BriEsh Telecom: Database-‐as-‐a-‐Service on HP Blades Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 12 Benefits 10x Faster Overall Greater Agility Deploy gold image Rapid, custom Lower Costs provisioning Weeks to minutes “What stands out … above the many other benefits of our Oracle Database consolidaAon onto a private cloud is the astonishing agility it gives us. Dropping our database deployment Ame down to 19 minutes from the weeks it used to take has brought us much closer to the customer service levels our business goals demand. Being able to manage 30% more databases with 20% fewer DBAs further confirms that standardizing on Oracle Database 11g was the right choice.’ Surren Partabh, CTO Core Technologies, BT Operate Business Objectives • Lower h/w and admin costs • Greater business agility • Better customer service Solution • Reduce h/w with DB consolidation • Standard configs: fast to deploy, easy to manage • Rapid self-service, user-customized ASM: 4 Common Pools of Storage • Data • Archive • Online Logs • Flashback Area:12 TB “Gold Image:” • 6 HP blades hos[ng 50 RAC DBs • Delivered in pairs: primary and DR • ReplicaEon with ODG Weeks to Minutes 20% less IT effort for 30% more DBs
- 13. Oracle RAC Customers – Strong on RAC Now and Then Con[nuing to go strong on Oracle Exadata Copyright © 2014, Oracle and/or its affiliates. All rights reserved. 13
- 14. CBA: Database-‐as-‐a-‐Service on “Commodity Clusters” ROI Rapid Copyright © 2014, Oracle and/or its affiliates. All rights reserved. Benefits “As the business seeks more services on demand, we needed to provide that resilient plaRorm and Exadata provides us a consolidated plaRorm for database services that can meet all those needs” -‐ Nicolas Tan, Head, Infrastructure and Plaiorm Services Business Objec[ves • Cost savings • Rapid provisioning • Extreme standardizaEon Solu[on • 2009-‐2013: Thirteen Exadata systems deployed in total 150% in 5 Years Provisioning Months to Minutes OPEX 50~60% Reduc[on Simplified Management Two Oracle Homes • Enterprise servers configured by CBA • Commodity clusters configured by CBA • Commodity clusters configured by Oracle • 350+ databases across 3 cloud “pools”
- 15. AmerisourceBergen: SAP ERP, CRM Benefits Copyright © 2014, Oracle and/or its affiliates. All rights reserved. “It (Exadata) allowed AmerisourceBergen to scale rapidly and double the size of the throughput at the company without interrupAon. The delivered soluAon is tremendously stable and scalable.” -‐ Philip Hotz, Manager EITS Oracle Support and SAP Basis Business Objec[ves • Increase capacity • Improve system stability • Scalable systems • Robust HA/DR • Reduce footprint Solu[on • 2013: 6 Full Racks + 3 Half Racks + Storage Expansion • 2014 plan: Half Rack upgrade + 1 Full Rack Data Center Cost Savings 2:1 Consolida[on Maximum Availability No unplanned down[me Scalability > 2x transac[ons +1 TB Data/Month Faster Time to Deploy Deploy in Days vs Weeks Exadata X3-‐8/X3-‐2 • SAP ERP, CRM, NetWeaver Portal, PI, IDM • Future: consolidate 100+DBs @ 230 TB+ • Full RMAN backup from standby • Archive log backup 1.5 TB/day Exadata X3-‐8/X3-‐2 + Storage Expansion • DR: 2 X3-‐8 full racks & 2 X3-‐2 half racks plus Storage Expansion half rack • Staging area for QA or UAT Pre-‐Exadata • IBM X64 Suse Linux • EMC storage arrays, SRDF replica[on • 9 systems per SAP applica[on Standby Produc[on WAN (60 miles) DR Data Guard Data Guard
- 16. Oracle Beehive: CollaboraEon Benefits Copyright © 2014, Oracle and/or its affiliates. All rights reserved. “Beehive is our largest applicaAon in-‐house. It is Oracle’s largest backend database.” -‐ Campbell Webb, Vice-‐President IT, Oracle Business Objec[ves • CollaboraEon for > 120K users • Headroom for 3+ years • Improved response Emes • Guaranteed upEme Solu[on • 2009: Move storage to Exadata V1 storage • 2011: Migrate to X2-‐2 Faster Response 5x – 60x Capacity for 100% Up[me Growth Exadata X2-‐2 Standby Exadata X2-‐2 Produc[on Exadata V1 Storage Servers AusEn (Texas) Data Center Utah Data Center 2011 • 9 full-‐rack X2-‐2 • 2.3 Petabytes raw disk • 48 TB flash • > 5,000 peak TPS • 9 full-‐rack X2-‐2 • Triple mirroring • Disk backups/flashback enabled • 100% up[me since go-‐live • 96 V1 storage servers • Post-‐Sun acquisi[on, CPU and disk oversubscribed Data Guard
- 17. Lower OpEx Increased Volume 43% less energy A Faster Applica[ons 2x – 7x faster Copyright © 2014, Oracle and/or its affiliates. All rights reserved. Rakuten: Real-‐Time OLTP Benefits Business Objec[ves • Support Rapid Business Growth • Reduce complexity and numbers of servers to manage • Balanced Workload and system Stability Solu[on • Replaced 6 HP SuperDomes and EMC storage with a single Exadata rack. MAA for DR. Replaced 6 HP SuperDome and EMC storage with Exadata 2x Exadata 5-‐nodes (primary and standby) • 2500 SQL transac[onsper second • 40 ms (avg) response [mes (99.99%) Maximum Availability 99.99% consumed. Data Centre cost reduc[on MAA is a framework. Availability must be a % Increase direcAve. ADG