130 likes | 134 Views
Designing and Implementing an Effective Data Warehouse Backup. Michael Goode Manager, Unix Administrative Services Media Technology Services Nielsen Media Research. Nielsen Media Research. Nielsen Media Research = TV Ratings
E N D
Designing and Implementing an Effective Data Warehouse Backup Michael Goode Manager, Unix Administrative Services Media Technology Services Nielsen Media Research
Nielsen Media Research • Nielsen Media Research = TV Ratings • NMR provides information on audience viewing habits for TV. In addition we provide information about computer usage and the Internet. • Headquartered in New York, New York Operations Center in Dunedin, Florida • Our customers are television programmers and commercial advertisers.
NMR Environment • Computers supported by Media Technology Services • 2 IBM Mainframes (OS/390 and VM) • 5 DEC servers (Alpha/VMS and VAX VMS) • 2 HP 3000 servers (MPE/iX) • 100+ Sun servers (Solaris) • 50+ Compaq servers (Novell and NT)
NMR Environment • Unix Administrative Services manages the Client/Server Backup Environment • Solaris Servers • Sybase Databases • Red Brick Warehouses • Solaris File Systems • NT Servers • Microsoft SQL Databases • Microsoft Exchange Databases • NT File Systems • Novell Servers • Novell File Systems
NMR NetBackup Environment • Unix Administrative Services manages the NetBackup Environment • Master Server - Sun Sparc 1000 (E5500 11/99) • 4 - STK RedWood Tape Drives • 2 - STK Timberline Tape Drives • Slave Server - Sun E6500 • 8 - STK 9840 Tape Drives • StorageTek Powderhorn Tape Silo • Capacity: 5300 Tapes
The Problem • A computer containing an 800 gigabyte (GB) Red Brick data warehouse had no backup and restore plan for the application. • My assignment was to design and implement a backup and restore solution for this application. Where do I begin?? • Hardware requirements? • NetBackup configuration issues? • Managing the backup process?
My Approach • Determine the hardware (tape drives) necessary to complete the backup and restore in the required time • Determine how to configure NetBackup to accommodate the warehouse file systems • Determine how to manage the backup process once in production
Tape Drive Selection • Determine the amount of data the warehouse will contain once in production • At time of request: 800 GB • At time of production: 1500 GB • Determine backup and restore windows • Negotiated a 6 to 8 hour window for backups • Restore must complete in one business day • 1500 GB / 6 hours = 250 GB/hour • Assess the current environment • 2 STK RedWood Tape Drives (80 GB/hour) • 1500 GB / 80 GB/hour = 19 hours
Tape Drive Selection • Consider Alternatives • Type of Tape Drives • Speed of Tape Drives • Floor Space Requirements • Choose Tape Drive Solution • 8 - STK 9840 Tape Drives (36 GB/hour each) • 1500 GB / 288 GB/hour = 5.25 hours • Test and Troubleshoot Solution • Calculations are theoretical, not practical • Test full restore capability too
NetBackup Class Definitions • Define NetBackup Classes • Multiplex? If so, how many? • Classes should be equally balanced • Review NetBackup Configuration • Maximum Jobs per Client? • NMR data warehouse (53 classes total) • Multiplex 5 classes per tape drive • Each class is approximately 38 GB • Maximum number of jobs per client = 40
Managing the Backup Process • Warehouse Backup Monitor Script • Starts 5 hours after warehouse backup • Runs once an hour • Generates email and file notification upon completion of the warehouse backup • Generates pager notification if the warehouse backup has not completed within 12 hours
Summary • This backup and restore solution met the requirements of the project team • Selecting the correct hardware (tape drives) was the most difficult aspect of this project • Do not forget to test the restore
Plans for the Future • Year 2000 Plans • Additional Slave Server to be used to create a High Availability NetBackup Environment • Rollout of Veritas HP3000 client • Evaluate Storage Manager, bpVault • Year 2001 Plans • Sun E10K Data Server • 5 - Red Brick Data Marts (1.2 TB each)