1 / 43

Improving Performance and Availability with Oracle-Specific Storage Management

Improving Performance and Availability with Oracle-Specific Storage Management. Melissa Stein VERITAS Software melissa.stein@veritas.com. Agenda. Challenges of Storage Management in Oracle environments Storage Management for Oracle Databases Overview Performance

rchadwick
Télécharger la présentation

Improving Performance and Availability with Oracle-Specific Storage Management

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Improving Performance and Availabilitywith Oracle-Specific Storage Management Melissa Stein VERITAS Software melissa.stein@veritas.com

  2. Agenda Challenges of Storage Management in Oracle environments Storage Management for Oracle Databases Overview • Performance • Continuous Data Availability Summary

  3. Challenges Every user needs high performance access to information Volume of data generated by each new major application has expanded exponentially Need to be assured that data is always available - 7x24x365 Disparate, non-integrated computing platforms need to share information

  4. • Voice/data integration • Wireless • Video-on-demand WEB - E-commerce Data Analysis S T O R A G E EnterpriseApplications Client/Server Basic Accounting Storage Fuels Applications- Databases Drive them 70’s 80’s 90’s 00’s 10’s

  5. S T O R A G E Storage Fuels Applications - Databases Drive them • Voice/data integration • Wireless • Video-on-demand WEB Data Analysis EnterpriseApplications Storage Management Curve Client/Server Basic Accounting 70’s 80’s 90’s 00’s 10’s

  6. Annual Brokerage System Cost Annual Internet Shopping Cost Annual downtime VAI 30 seconds $53,750 $950 5 minutes $537,500 $9,417 52 minutes $5,590,000 $98,000 8.75 hours $56 million $988,750 A V A I L A B I L I T Y 43.7 hours $280 million $5 million 87.6 hours $560 million $10 million 180+ hours $1+ billion $20+ million 450+ hours $3+ billion $50+ million % uptime 99.9999 99.999 99.99 99.9 99.5 99.0 98.0 95.0 Savings ($)

  7. Storage Management Objectives Configuring and maintaining Storage Resources to: • Maximize I/O performance • Is I/O done in an optimal manner for my application? • Is my Storage configured optimally? • Minimize Loss of Access to Data: “No time for down time” • Meet service level agreements • Optimize utilization of resources

  8. Overview of Storage Management Components Storage Mgt Tools Web Server Backup Application Application Level HSM ORACLE File System Volume Mgt Host OS Driver Storage Subsystem Storage Subsystem

  9. Storage is the Heart of a Database System • Database functions to: • store, update, and retrieve structured information • Good storage design helps: • Meet rigorous performance and availability requirements of production databases

  10. Performance

  11. Performance can be Everything • Need to meet service level agreements • Elegant designs can fail in production because of performance problems • Administrators spend hours monitoring and balancing I/O loads to achieve better performance

  12. VERITAS Database Edition as a Highly Performing Platform for Databases • Best of both worlds: Raw I/O performance and ease of file system administration Oracle Raw Disks Conventional File Systems Advantages Disadvantages Advantages Disadvantages • Manageability • Tablespace • organization • System-wide • backup and • restore • policies • Easy to create • and expand • files for Oracle • Manageability • “Invisible” • locations for • tablespaces • Difficult to • grow • tablespaces • Performance • Lower database • throughput • (30-90% of • raw disk) • Higher CPU • overhead • Data Reliability • Potential data loss • Performance • Best OLTP performance • Data • Reliability • No potential for data loss with data written to • disks directly

  13. Quick I/O Features • Direct (non-buffered I/O) • Buffered reads in case of Cached Quick I/O • Bypasses the file system layer’s locking • Supports asynchronous I/O • Easy Online Management to turn Cache on/off for Quick I/O files • Multiple Database instances can co-exist with selective caching for specific data files • Minimum changes to existing configuration • No change to the OS or Database version levels

  14. Quick I/O raw device regular file Raw Device PerformanceDatabase Edition’s Quick I/O • Break-through VERITAS File System interface technology available through VERITAS Database Edition • Presents regular VERITAS File System files to Oracle as raw character devices • Allows parallel updates to database files for increased throughput • Takes advantage of kernel async I/O (Solaris’ KAIO) • Oracle handles locking for data integrity • Eliminates traditional UNIX file system overhead RDBMS Server

  15. 8000.00 7000.00 6000.00 5000.00 tpm 4000.00 3000.00 2000.00 1000.00 0.00 250 MB 500 MB 1 GB 1.5 GB 2 GB Oracle Buffer Size Raw Partition Database Edition with Quick I/O Database Edition with Cached Quick I/O UFS with buffered I/O (default) UFS with direct I/O Quick I/O Performance

  16. Clients Clients Network Secondary Server Primary Server Replication Database Replication OptionIncreases Performance • Off-load processing of maintenance tasks to secondary server • Backup or decision support can then be performed on data on second system • Low CPU and network usage due to only replicating changed data blocks Higher Performance on Primary Server

  17. Disk Block Disk Block Disk Block Disk Block Disk Block Disk Block Disk Block Disk Block Disk Block Disk Block Disk Block Disk Block Veritas File Systm Inode Disk Block Disk Block Extent Addresses Disk Block Disk Block Direct Address 0 / Length Direct Address 1 / Length Direct Address 2 / Length Direct Address 3 / Length Direct Address 4 / Length Direct Address 5 / Length Direct Address 6 / Length Direct Address 7 / Length Direct Address 8 / Length Direct Address 9 / Length Single Indirect Block Double Indirect Block Increased Performance for Large I/O • Extent-based allocation can accelerate I/O by increasing the I/O transfer size in the File System • Optimal for multimedia files commonly found on the Internet

  18. File Copy Pt. 1 File Pt. 2 File Pt. 3 File Pt. 1 Copy Pt. 2 Copy Pt. 3 Copy Pt. 1 File Pt. 2 File Pt. 3 File Parity RAID Functionality RAID - 1 or Mirroring: Exact duplicate volumes. Full Redundancy. RAID - 0 or Striping: Stripes data across all drives. No Redundancy. RAID - 0 + 1 or Mirrored Stripes: Stripes data across all drives. Mirror with another RAID 0 column. Full Redundancy with Full Performance. RAID - 5: Stripes data across all drives with parity. Full Redundancy. Doesn’t require duplicate storage investment.

  19. Col 1 Col 2 Col 3 Col 1 Col 2 Col 3 Increase Volume Size Col 1 Col 2 Col 3 Col 4 Volume Manager Increases PerformanceThrough Proper Storage Configuration • RAID functionality without hardware • Minimization of I/O bottlenecks • Maximum throughput &bandwidth of I/O • Online Relayout • Ability to change configuration ONLINE • Enables volume growth by adding columns • Incremental growth • Ability to convert a RAID5 volume to a striped volume ONLINE • Allows ONLINE changing of stripe unit size

  20. Continuous Data Availability

  21. Business Without Interruption Database EnvironmentsIncreasingly Complex Increased Risk toData & Bottom Line Storage Management SoftwareHelps Mitigate Risk

  22. Lost Revenue Associated With Application Outages Application Call location Number portability Enterprise resource planning Supply chain mgmt Electronic commerce Internet banking Universal phone services Customer service center POS/EFT Messaging Cost Per Min $27,000 $14,000 $13,000 $11,000 $10,000 $7,000 $6,000 $3,700 $3,500 $1,000 Costs of Information • Cost of downtime is the easiest way to justify availability • Deploy on existing hardware and/or software • If using Service Level Agreements (SLA) then clusters are the way to meet them IDC April ‘99

  23. Online Storage Management • Perform all disk & file management operationswhile applications are running • Volume Management • Move data between devices for load balancing • Reorganize storage layout • Dynamic Multi-Pathing • Hot sparing • Hot swapping • Fast recovery mechanisms • File System • Journaling / fast recovery • Online growth & shrinkage • Online reorganization / defragmentation

  24. Disk 1 Disk 5 Disk 1 Disk 5 Disk 5 Disk 1 Disk 1 Redundancy Disk 6 Disk 2 Disk 6 Disk 6 Disk 2 Disk 2 Disk 2 Disk 3 Disk 3 Disk 7 Disk 7 Disk 3 Disk 3 Disk 7 Redundancy Disk 4 Disk 8 Disk 4 Disk 8 Disk 8 Disk 4 Disk 4 Mirror 1 Mirror 1 Mirror 1 Mirror 2 Mirror 2 Disk 5 Disk 6 Mirror 1 Mirror 2 Disk 7 Disk 8 Mirror 2 RAID Redundancy Enhanced Redundancy • Increases ability to handle multiple disk failures • Reduces amount of resynchronization • Keeps database data available

  25. High Availability • Why? • Service resumes with minimal impact • Administrator fixes problem in own time rather than with users shouting!

  26. Cluster: a type of parallel or distributed system that: consists of a collection of interconnected whole computers used as a single, unified computing resources Clustering (VCS) Resource Groups Clients Servers Shared Storage RG-Web Node A RG-NFS Node B Win98 RG-Oracle switch Node C RG-SAP Node D Client -1 RG-NBU Pfister - In Search of Clusters

  27. GeoCluster • Monitor and control of multi-heterogeneous clusters at multiple sites • Disaster recovery between sites • Ability to trigger action on a cluster based on events on other clusters • Follow-the-sun migration DR

  28. Replication (SRVM) • Replicate the global environment over a Wide area network • Offers disaster recovery solution

  29. Increased Availability throughBlock-Level Incremental Backup • Back up only changed data blocks! • Increased data availability • Virtually eliminates backup windows • Allows more frequent backup schedules • Backups contain up-to-date data • Faster recovery from backups (vs. applying changes in logs) • Integration of NetBackup with Database Edition

  30. + Block-Level Incremental Backup NetBackup Server Changed BlocksSince LastBackup VERITAS File System with Storage Checkpoints NetBackup BLI BackupExtension Minimize Your Database Backup Window PreviousFull Backup BlockLevelIncrementals FullRestore

  31. Faster Database Recovery through Storage Rollback Recovery from Storage Checkpoints = Simply write back changed (blue) blocks to previous state and then apply redo logs to bring database to desired point in time

  32. Performance Availability Storage Management Integration of Backup/Recovery, Disk & File Management, Databases, and Hardware Subsystemsto Increase of Oracle Databases

  33. How to Configure Oracle/VERITAS/EMC Paul Tsien, Marty Ward, Garry Lemasa Oracle, VERITAS, EMC

  34. Goals • Provide configuration recommendation for using Oracle/VERITAS/EMC • Focus on simplicity and manageability • Assumption - RAID 0+1 • High availability • Performance is important

  35. The Challenges • Complexities • Configuring Oracle, VERITAS, or EMC • Dependencies and trade off • H/W vs. S/W striping, cache effects, etc. • Maintain availability and performance in a dynamic environment • Adding storage, application changes, etc. • Requires monitoring and re-tuning

  36. The Traditional Configuration (Striping) Approach • Layout database files on storage based on: • Application types • OLTP vs. DSS • Database file I/O characteristics • Log, data, temp, etc. • Maximum flexibility for performance tuning • Custom configuration (complex)

  37. Technology Improvements Make New Approach Possible • VERITAS • Efficient and scalable striping of files • Efficient file system I/Os using Quick I/O • Dynamic data re-configuration (striping) • EMC • Intelligent cache (up to 16GB) • Highly reliable mirrored disk with online disk replacement • Dynamic load balancing • Oracle • Parallel database execution

  38. The New (Striping) Approach • Stripe and mirror everything • Mirroring protects data • Striping reduces hot spots (balance the load) • Simple and easy • Will it work?

  39. The Environment • System • Sun E3000 4x248MHz CPUs, 2GB memory, 1x64-bit Fibre Channel HBA • Storage • EMC Symmetrix 3630; 32x36GB, 2GB cache • Software • Solaris 2.6 • VERITAS Database Edition for Oracle 2.0.2 • Oracle Database • Oracle8i (8.1.6) • 100GB Account Receivables

  40. Test Configurations Stripe Set Stripe Depth 1 2 31 32 Slice 33 •• • • • • •• M33 M1-M33 are mirror copies M32 M31 M2 M1 Disk 1 Disk 2 Disk 31 Disk 32 Disk 1 = 36GB (4 slices of 9GB) Slice or Hypervolume = 9GB Stripe Set = 32 Stripe Width = 33 Stripe depth = 64K, 256K, 1MB

  41. Data, Log, Temp, etc. •• • • • • •• Mirror Disk 1 Disk 2 Disk 31 Disk 32 Test Configurations Config #1 Data, Log, Temp, etc. Config #2 •• • • • • •• Mirror Disk 1 Disk 2 Disk 31 Disk 32 Data Log Config #3 •• • • • • •• Mirror Mirror Disk 1 Disk 2 Disk 31 Disk 32 Disk 1 = 36GB (4 slices of 9GB) Slice or Hypervolume = 9GB Stripe Set = 32 Stripe Width = 32, 16 Stripe depth = 64K, 256K, 1MB

  42. Preliminary Results • Avoid disk head contention • Do not put data and mirror on same physical disk • Stripe and mirror everything works • Improved sequential performance • Good mix load performance • More consistent performance numbers • VERITAS Quick I/O works • Improved performance significantly • Factors to consider - availability and good performance given ($$$) • Manageability, last 5% performance, flexibility

  43. Business without interruption. Thank you!

More Related