Posted by: Anonymous Coward
on June 15, 2005 04:01 AM
Sorry to go OT. Yes, sessions are handled fine if you configure the load balancer to send the same client connect back to the same cluster node this is not a problem. With LVS there are lots of ways to do this (called LVS persistence).
As for the performance bottleneck on the back-end session DB: you needn't put all your eggs in one basket (or do you?). If you don't need to worry about highly available sessions (session restarts are okay if a node goes down) and you don't need to worry about locking issues why not store session information local to each cluster node? When the user selects items for the shopping cart you have to reduce the item from inventory on the back-end DB but maybe you can have different databases running on different backend servers (toys, housewares, garden, etc. each on their own DB server). Of course, if you really are hitting the wall and you can't build your application around a physical architecture then you are probably a big enough player to look at distributed cluster file systems that will eliminate performance bottlenecks--something like Polyserve (www.polyserve.com). You can put a Polyserve cluster for the SQL servers that sits behind the LVS load balanced cluster of application servers.
Re:How does this work with sessions?
Posted by: Anonymous Coward on June 15, 2005 04:01 AMAs for the performance bottleneck on the back-end session DB: you needn't put all your eggs in one basket (or do you?). If you don't need to worry about highly available sessions (session restarts are okay if a node goes down) and you don't need to worry about locking issues why not store session information local to each cluster node? When the user selects items for the shopping cart you have to reduce the item from inventory on the back-end DB but maybe you can have different databases running on different backend servers (toys, housewares, garden, etc. each on their own DB server). Of course, if you really are hitting the wall and you can't build your application around a physical architecture then you are probably a big enough player to look at distributed cluster file systems that will eliminate performance bottlenecks--something like Polyserve (www.polyserve.com). You can put a Polyserve cluster for the SQL servers that sits behind the LVS load balanced cluster of application servers.
#