BukManiac/Scaling the Database
CS290F Fall 2006 - UCSB Computer Science - Thorsten von Eicken
BukManiac -- Your Search for Books ends here !
Home | Overview | Design | Critical Paths & Analysis | Optimizations | Scaling the Database | Scaling the Application | Plugging in the Cache
Contents |
Scaling the Database
We started out with a database having 10,000 book records, then scaled it up to 104,000 records. But we still thought it was fitting into the main memory. Then, we populated the database to have 270,000 book records. We observed interesting behavior.
Effect on Reply Rate
This graph shows the effect of increasing the DB size. We observe an average of 15% decrease in replies that the server returns. This is primarily because database has to access the disk frequently and page data in and out of memory
Effect on Response Time
The response time also increase as an effect of increasing the database size. The time to access the database increases since the DB engine has to go to the disk to get results.
Further possible Improvements
After scaling on 1 server, we can further scale our application on multiple servers. After this, our next step would be to use caching to minimise the number of hits to the database for the search results.
