On Tue, Apr 29, 2008 at 9:16 AM, Richard Eng richard.eng@rogers.com wrote:

Just out of curiosity, if I should need to scale my Seaside app beyond my one Linux box, exactly how do I do that?

I am currently using Apache to load-balance over several Squeak VMs:

<Proxy balancer://seaside_cluster> BalancerMember http://localhost:9090/seaside/go route=GSN9090 BalancerMember http://localhost:9091/seaside/go route=GSN9091 BalancerMember http://localhost:9092/seaside/go route=GSN9092 BalancerMember http://localhost:9093/seaside/go route=GSN9093 ...

Do I add a second Linux box that runs additional Squeak VMs (but does not run Apache)? Say, the second Linux box is at IP 192.168.1.110. Then do I edit the above to:

<Proxy balancer://seaside_cluster> BalancerMember http://localhost:9090/seaside/go route=GSN9090 BalancerMember http://localhost:9091/seaside/go route=GSN9091 BalancerMember http://localhost:9092/seaside/go route=GSN9092 BalancerMember http://localhost:9093/seaside/go route=GSN9093 ... BalancerMember http://192.168.1.110:9200/seaside/go route=GSN9200 BalancerMember http://192.168.1.110:9201/seaside/go route=GSN9201 BalancerMember http://192.168.1.110:9202/seaside/go route=GSN9202 ... ?

Is it really that simple and straightforward?

Yes it is. What is more, you can setup a net like this:

(Internet) --> (dedicated proxy/balancer or hardware proxy/balancer) --> (a lot of boxes running a lot of images of your app)

This way you can push all your static content to the dedicated proxy so it can be cached in memory for faster serving. Also, you can put your certificate in this box so all the content is served to the internet encrypted. The lot of boxes are just normal servers without certificates or static content, just your simple and plain app. All the cpu cycles available for your app.

Also, you can add/remove nodes in the farm of servers with the image, for example for adding or removing capacity (also it can be Amazon EC2 images added/removed on demand) and just configure your proxy (although most proxys/load balancers detect nodes lost and automatically avoid sending more sessions to them.

Also, have in mind that you need a sticky session capable load balancer. You need this as the callbacks live in the image and are not capable of migrate to new servers, so the same session has to be routed to the same server where it was created.

Miguel Cobá

Thanks, Richard _______________________________________________ seaside mailing list seaside@lists.squeakfoundation.org http://lists.squeakfoundation.org/cgi-bin/mailman/listinfo/seaside