5๐
โ
You must be able to handle N concurrent queries (N โ concurrent new
users).
You see the difference at a very large scale only. You must be pessimist (server is down, bad buckup, amazon bancrupcy ) and your database must be able to handle all users in rush hours.
- 10 000 000 users with cookie session will cost you 0$ extra, problems with iframes and mobiles
- 10 000 000 users with cache session will cost you hundreds or less than hundred of dollars daily, lost sessions in case of restart
- 10 000 000 users with cache_db session will cost you thousands of dollars daily (any persistent storage able to handle 1 000 000 hits in one time, 10% of users)
I use redis backend (periodic save to disk) with the possibility of switching to a cluster of memcached services.
http://www.icis.com/blogs/icis-chemicals-confidential/files/2011/10/19/goodcheapfast.jpg
๐คbaklarz2048
Source:stackexchange.com