With CFMX Enterprise, you can leverage J2EE Sessions, that will allow you to duplicate sessions across a cluster.
Sessions have the advantage of being able to store complex data types, and don't involve a call to a database.
However, I've seen CF apps with client variables in a database in a clustered non-sticky environment work well.
I would lean towards the J2EE duplicating sessions as a solution personally. You can read more about it at www.bpurcell.org.
- Calvin
-----Original Message-----
From: Al Everett
Date: 9/29/04 9:51 am
To: CF-Talk
Subj: SESSION vs. CLIENT
We're working in a clustered environment (eight servers in one pool, two
in several others) on Windows 2000/IIS 5.0 behind a BigIP load balancer
with some CF5 and some CFMX. (Very soon to go all CFMX.)
We use Session variables extensively to keep track of user's logon
information, which has required us to use "stick sessions" on the load
balancer to keep users stuck to a server. However, we've gone through
several exercises with certain customers because of the way their proxy
servers/firewall block or trap cookies or other issues.
I have suggested that if we were to go to Client variables in a database
for that information we would remove the need for sticky sessions. The
argument against me has always been that a database transaction is too
expensive. Back before our database was tuned and we resolved a bunch of
network issues I would have agreed, but we're almost guaranteed to have
at least one call to the database on every page load right now and the
site is fine. (Oracle 9i running on Solaris.)
Does anybody have any good statistical and/or anecdotal evidence on
whether Sesssion is better or worse than Client variables, specifically
in CFMX6.1?
_______________________________
Do you Yahoo!?
Express yourself with Y! Messenger! Free. Download now.
http://messenger.yahoo.com
[Todays Threads] [This Message] [Subscription] [Fast Unsubscribe] [User Settings] [Donations and Support]
