Jun
11
2008
Is Anyone Out There Still Using ClusterCats?
Posted by Mike Brunt at 12:57 PM
7 comments - Categories:
ColdFusion
Conversations With Mike Brunt About ColdFusion Server Side Stories
Comments
Allen wrote on 06/11/08 3:03 PM
Cluster cats? Nw there's a name I haven't heard for a long time!
Mike Brunt wrote on 06/11/08 8:47 PM
@Mark Of course I'm insane ;o) well slightly mad because the current iteration of Clustering we are using in ColdFusion is very limited. Mainly because we are using what is available in J2EE and it is purely peer to peer.ClusterCATS was a very full featured piece of software and of course it was a challenge to get up and running; very network-TCP/IP intensive. I was lucky enough to work with some of the senior engineers who created the software at the company Allaire purchased to create ClusterCATS - Bright Tiger.
@Allen sorry to plethor you both with minutia.
tony petruzzi wrote on 06/12/08 5:15 AM
@mike,what advantage did cluster cats offer over just over a hardware load balancer?
i remember people talking about it back in the 4-5 days. i never really saw the advantage of using it.
personally whenever i needed to build a "clustering" site, i would avoid session variables and just use cookies and client variables. then i would just have a hardware load balancer on the from end of the farm. worked like a champ and no headaches.
Ryan TJ wrote on 06/12/08 9:02 AM
Never used CC but evaluated it, sure it was ok but the hardware load balancers offered better networking type features, not needing a dns entry for each server, ssl acceleration... We use Cisco CSS and F5 has nice products also. Frankly the only feature I like about Jrun clustering is if we needed to have the session replicated we could but today we would rebuild from the db if the user failed over to another server. If we did use Jrun we'd probably combined the hardware cluster to point at jrun clusters of two servers.
Mike Brunt wrote on 06/12/08 10:15 AM
@tony ClusterCATS big advantage at the time was that it was much more cost-effective than hardware Clustering. In some ways it still would be, in this sense. I believe any web site used by the public needs to have some redundancy, it needs High Availability (HA) of some kind so fully effective software level Clustering still has a place for those who still cannot afford a hardware Clustering device either in purchase terms or ongoing administration costs.
Mike Brunt wrote on 06/12/08 10:20 AM
@Ryan, the comments I made in my response to Tony's thoughts apply also to your points. In addition, I have assisted many clients in using the J2EE session replication that JRun uses. From what I have seen, I would avoid using in-memory session state at all. Client variables would be my ideal choice and I would look very carefully at the need for complex data types and try to architect my database to remove that need from the code-level.I hope this makes sense and thanks to both of you for your comments.
Mark Drew wrote on 06/11/08 2:39 PM
Are you insane??! When I stopped using it, I actually started living. I lost hair and went grey using that! I am not even joking!