I am planning to use HBase for a social application. The starting user base would be around 5k and monthly it would keep increasing on around some 3k per month.
Which approach would increase the efficiency of my application:
- Dynamically adding resources on the go (I am thinking the read would suffer if i am trying to load data across months).
- Adding static resources, in the initiation of the application (which can handle all the 25TB of data with appropriate region server and region splits for uniform writing of data and reading would be faster).
P.S.: The average data which my application would be handling over an year would amount to ~25TB. According to HBase Capacity Planning and Region Sizing, the application would be needing 44 region servers(without needing to change the default configurations).
Any suggestions?
