Why did you edit this back? Scanning 1T in 3.5 seconds is impossible with the current hard drive and interface technology. Each Exadata cell does work that the database machine would normally do. This alleviates the load on the database machine. However, Exadata still uses plain old hard drives. Each hard drive can do a MAX of 100MB/sec. That is *megabytes* a second. Since there are 12 disks, you can expect that reading from all hard drives simultaneously would give a throughput of 1200 MB/sec. In that case, scanning 1TB would take 1,000,000MB / 1200MB = 833 seconds, or 13 minutes.
- If I remember correctly, the 3.5 seconds was mentioned by Larry Ellison himself while announcing the Exadata at OOW2008. Anyway, using your calculation (assuming for now that it's correct): 12 disks per cell and cluster 6 cells together, gives you 72 disks per system. So, scanning 1TB would take 1,000,000MB / 7200MB = 138.89 seconds, or 2.3 minutes. Best regards. Frank Naude 05:08, 7 January 2009 (UTC)