I am working on a 'back of the envelope' calculation for a 100TB reporting database setup. I am seeking out thoughts from the experts here. Proposed environment:
- Storage Capacity ~ 100TB
- Tables ~ 200, sizes ranging from 1GB to 5TB. mean size could lie between 100GB-200GB
- ETL - jobs may require join between tables of 10's of millions of rows, with join keys ranging from 10 bytes to 500 bytes. such joins should finish in under 2-5 minutes
- Live Selects - initially, only interested in select speeds. should support 500 selects/second. Updates/second will be relatively much smaller number and can be ignored for this exercise.
- need 24x7 availability. 2 independent DB servers should be available to serve select calls (with data replicated).
- At present, I am looking at Oracle. How has your experience been with other commercial (or) opensource solutions for large databases?
- What hardware-OS have you seen to work best? I am planning for Linux on Dell.
- Is Network storage, such as NetApp, a must? What issues do you foresee with using commercial off the shelf disks?
- Once the hardware and OS are ready, how much time would you set aside to setup, configure DB, storage etc.
- What team compositions worked best in the environments you have observed? I mean, the various Admins (OS Admin, Oracle DB Admin?) required to manage and operate such a setup. How many of them might be needed to achieve a 24x7 uptime.
- Any approximation/range on DB Licenses, Network Storage costs.
I know I dont have all the environment details. I am not looking for exact details, an approximation is sufficient. Though some of the questions might be best answered by managers, I am interested in Admins perspective. I appreciate your input.